na_experiments.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YqeiKwnV4ms9"
      },
      "source": [
        "Teemu Sidoroff (teemu.sidoroff@helsinki.fi)            2023-05-16\n",
        "\n",
        "---\n",
        "\n",
        "\n",
        "# Number agreement in BERT\n",
        "\n",
        "## Introduction\n",
        "\n",
        "This project looks into the syntactical capabilities of pre-trained BERT models. The project reproduces the \n",
        "experiments of subject verb number agreement tasks. \n",
        "\n",
        "The experiment is based on the material and prior research by Yoav Goldberg ([2019](https://arxiv.org/abs/1901.05287)) which studies\n",
        "how BERT captures the subject-verb agreement by masking the verb in question and evaluating how accurately\n",
        "the model selects the verb with correct form (plural vs singular) in the masked word prediction. As an addition to Goldberg's work my \n",
        "aim is to apply the same method by inverting the task by masking the subject word instead of the verb and compare the \n",
        "results between the two.\n",
        "\n",
        "This Google Colab notebook and the associated data is available at [Github repository](https://github.com/tsido/LDA-T307-project).\n",
        "\n",
        "## Datasets \n",
        "\n",
        "Goldberg investigated the task using three different data sets. I use two of these datasets and create\n",
        "another version of the sets where the subject noun is masked instead of the verb. This allows the comparison of \n",
        "the performance of the model between the two tasks.\n",
        "\n",
        "The first dataset consists of around 30,000 sentences. In addition to the sentences and information of the masked verb the dataset includes information of how many attractors (i.e. other nouns between the subject and the verb) the sentences contains. The number of attractors ranges from 1 to 4 in the dataset. This dataset is referred as the large dataset in this notebook. \n",
        "\n",
        "The second dataset is a smaller one, consisting of nonsensical but syntactically correct sentences with 0 to 2 attractors in each. This dataset is considerably smaller, consisting only of 380 sentences. The idea of using these nonce sentences is to avoid the model from using semantic cues, but rely on syntactical information for the task instead. The dataset is referred as nonce dataset in the following.\n",
        "\n",
        "## Method\n",
        "\n",
        "To generate the subject masked versions of the datasets, I use [spaCy](https://spacy.io/) library to identify the subject for the sentences. This parsing step can introduce some bias into the data, as the\n",
        "parser fails to parse all sentences correctly. For generating\n",
        "the plural or singular version of the noun I used the [pattern](https://www.digiasset.org/html/pattern.html) library. \n",
        "\n",
        "I limited the study to investigate the results of only a single model, namely the [BERT base model](https://huggingface.co/bert-base-cased). \n",
        "\n",
        "The following section of this notebook contain the experiment code that can be replayed by re-running this notebook.\n",
        "\n",
        "The results of the experiment are discussed below in the following sections.\n",
        "* [Results for the large dataset](#large-dataset-results)\n",
        "* [Results for the nonce dataset](#nonce-results)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Experiment code\n",
        "\n"
      ],
      "metadata": {
        "id": "Jxz1d3A4KAlT"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "RADHc5H-44kI"
      },
      "outputs": [],
      "source": [
        "# avoid spam\n",
        "%%capture\n",
        "\n",
        "!pip install transformers\n",
        "!pip install pattern\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 2,
      "metadata": {
        "id": "Le1bWZuR4ms7"
      },
      "outputs": [],
      "source": [
        "import pandas as pd\n",
        "from transformers import PreTrainedModel, PreTrainedTokenizer,BertTokenizer, BertModel,BertForMaskedLM\n",
        "from transformers import pipeline\n",
        "import torch\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 3,
      "metadata": {
        "id": "HJom2v2W9r-d"
      },
      "outputs": [],
      "source": [
        "%%capture\n",
        "!mkdir data\n",
        "# These are copies of the datasets by Goldberg, also available at https://github.com/yoavg/bert-syntax\n",
        "!wget --output-document data/generated.tab https://raw.githubusercontent.com/tsido/LDA-T307-project/main/data/generated.tab\n",
        "!wget --output-document data/lgd_dataset.tsv https://raw.githubusercontent.com/tsido/LDA-T307-project/main/data/lgd_dataset.tsv\n",
        "!wget --output-document data/nonce_dataset.csv https://raw.githubusercontent.com/tsido/LDA-T307-project/main/data/nonce_dataset.csv"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7b8Pp_h04ms-",
        "outputId": "e90ee6ef-d40b-45c0-c6c3-1175a5e17f6e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Some weights of the model checkpoint at bert-base-cased were not used when initializing BertForMaskedLM: ['cls.seq_relationship.bias', 'cls.seq_relationship.weight']\n",
            "- This IS expected if you are initializing BertForMaskedLM from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
            "- This IS NOT expected if you are initializing BertForMaskedLM from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
          ]
        }
      ],
      "source": [
        "%%capture\n",
        "# Initialize the model and tokenizer utlized in the experiments\n",
        "model = BertForMaskedLM.from_pretrained('bert-base-cased')\n",
        "tokenizer = BertTokenizer.from_pretrained('bert-base-cased')\n",
        "\n",
        "device = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n",
        "model.to(device)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 191
        },
        "id": "nuJOmEHz4mtA",
        "outputId": "8bdb765c-ffd4-471e-d5ff-fc8f46bdfc5e"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "       attractors                                      full_sentence  \\\n",
              "14548           1  other landmarks in the town include our lady o...   \n",
              "10090           2  it has been observed that cells containing a s...   \n",
              "\n",
              "                                    verb_masked_sentence verb_true_label  \\\n",
              "14548  other landmarks in the town [MASK] our lady of...         include   \n",
              "10090  it has been observed that cells containing a s...            tend   \n",
              "\n",
              "      verb_false_label  \n",
              "14548         includes  \n",
              "10090            tends  "
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-87fb793b-e6b0-4c8e-b5ab-6fff351b456e\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>attractors</th>\n",
              "      <th>full_sentence</th>\n",
              "      <th>verb_masked_sentence</th>\n",
              "      <th>verb_true_label</th>\n",
              "      <th>verb_false_label</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>14548</th>\n",
              "      <td>1</td>\n",
              "      <td>other landmarks in the town include our lady o...</td>\n",
              "      <td>other landmarks in the town [MASK] our lady of...</td>\n",
              "      <td>include</td>\n",
              "      <td>includes</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>10090</th>\n",
              "      <td>2</td>\n",
              "      <td>it has been observed that cells containing a s...</td>\n",
              "      <td>it has been observed that cells containing a s...</td>\n",
              "      <td>tend</td>\n",
              "      <td>tends</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-87fb793b-e6b0-4c8e-b5ab-6fff351b456e')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-87fb793b-e6b0-4c8e-b5ab-6fff351b456e button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-87fb793b-e6b0-4c8e-b5ab-6fff351b456e');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 5
        }
      ],
      "source": [
        "# Linzen Goldberg dataset (=large dataset): ~30k sentences with 1-4 attractors\n",
        "df = pd.read_table('data/lgd_dataset.tsv', header=None, names=['attractors', 'full_sentence', 'verb_masked_sentence', 'verb_true_label', 'verb_false_label'])\n",
        "\n",
        "df['verb_masked_sentence'] = df['verb_masked_sentence'].apply(lambda x : x.replace(\"***mask***\", \"[MASK]\"))\n",
        "df[['attractors','full_sentence', 'verb_masked_sentence', 'verb_true_label','verb_false_label']].sample(2)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "id": "GYijcdOO4mtA"
      },
      "outputs": [],
      "source": [
        "import math\n",
        "\n",
        "# the dataframe is called sample_df since during implementation\n",
        "# I used a small subset of the data. Leaving the naming as is even\n",
        "# when using the full data set\n",
        "# Running with 30k row dataset and GPU this takes some 8 minutes in Google Colab\n",
        "#sample_df = df.sample(300)\n",
        "sample_df = df\n",
        "\n",
        "# extract the logits for the singular and plural version \n",
        "# for the word in the masked position of a sentence\n",
        "def get_label_probs(row, pos):\n",
        "  if (pos == 'verb'):\n",
        "    masked_sentence = row['verb_masked_sentence'] \n",
        "    true_label = row['verb_true_label']\n",
        "    false_label = row['verb_false_label']\n",
        "  else:\n",
        "    masked_sentence = row['subject_masked_sentence'] \n",
        "    true_label = row['subject_true_label']\n",
        "    false_label = row['subject_false_label']\n",
        "\n",
        "\n",
        "  # skip sentences we failed to parse (happens in the noun masking case) \n",
        "  if (masked_sentence == None):\n",
        "    return (float('nan'), float('nan'), float('nan'))\n",
        "\n",
        "  # tokenize the text\n",
        "  encoded_input = tokenizer(masked_sentence, return_tensors='pt').to(device)\n",
        "    \n",
        "  output = model(**encoded_input)\n",
        "\n",
        "  # extract the index for the [MASK] token\n",
        "  input_ids = encoded_input['input_ids'].squeeze().tolist()\n",
        "  mask_index = input_ids.index(tokenizer.mask_token_id)\n",
        "\n",
        "  true_label_id, false_label_id=tokenizer.convert_tokens_to_ids([true_label, false_label])\n",
        "  true_label_p = output[0][0][mask_index][true_label_id].item()\n",
        "  false_label_p = output[0][0][mask_index][false_label_id].item()\n",
        "\n",
        "  return (true_label_p, false_label_p, true_label_p > false_label_p)\n",
        "\n",
        "# prefix results with 'verb_' for the masked verb experiment\n",
        "sample_df[['verb_true_label_p', 'verb_false_label_p', 'verb_is_accurate']] = sample_df.apply(lambda x: get_label_probs(x, 'verb'), axis=1, result_type='expand')\n",
        "\n",
        "sample_df.to_csv('verb_number_agreement_with_results.csv')\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "id": "EqDyU9RQ4mtB"
      },
      "outputs": [],
      "source": [
        "\n",
        "# create versions of the sentences where the **subject** is masked and re-create the experiment\n",
        "\n",
        "# To do the subject masking, do the following:\n",
        "# - parse the full sentences with spacy and identify the first subject\n",
        "# - create subject masked version of the sentence and do the same comparison as with the verb case\n",
        "\n",
        "import os\n",
        "import spacy\n",
        "from pattern.en import pluralize, singularize\n",
        "\n",
        "SUBJECT_NOT_FOUND_FILE = \"subject_not_found_dump.csv\"\n",
        "DEBUG = False\n",
        "\n",
        "nlp = spacy.load('en_core_web_sm')\n",
        "\n",
        "def get_subject(sentence):\n",
        "    doc = nlp(sentence)\n",
        "    for token in doc:\n",
        "        if token.dep_ == 'nsubj':\n",
        "            return token.text\n",
        "    return None\n",
        "\n",
        "\n",
        "def isplural(word):\n",
        "     singular = singularize(word)\n",
        "     return False if word == singular else True\n",
        "\n",
        "# return the masked sentence and the subject and the singular and plural subjects\n",
        "def generate_subject_data(row):\n",
        "    sentence = row['full_sentence']\n",
        "    subject = get_subject(sentence)\n",
        "    if (subject):\n",
        "        masked_sentence = sentence.replace(subject, '[MASK]')\n",
        "\n",
        "        singular = singularize(subject) if (isplural(subject)) else subject\n",
        "        plural = pluralize(subject) if (not isplural(subject)) else subject\n",
        "\n",
        "        false_label = plural if singular == subject else singular\n",
        "        true_label = singular if singular == subject else plural\n",
        "\n",
        "        return (subject, masked_sentence, true_label, false_label)\n",
        "    else:\n",
        "        if DEBUG: print(\"No subject found:\", sentence)\n",
        "        #store the whole row into a separate file for manual inspection\n",
        "        row.to_frame().T.to_csv(SUBJECT_NOT_FOUND_FILE, mode='a', header=False, index=False)\n",
        "        return (None, None, None, None)     \n",
        "\n",
        "\n",
        "# write the discarded sentences into a file, remove existing file first\n",
        "# for later inspection\n",
        "if os.path.exists(SUBJECT_NOT_FOUND_FILE):\n",
        "    os.remove(SUBJECT_NOT_FOUND_FILE)\n",
        "empty_df = pd.DataFrame(columns=df.columns)\n",
        "# write the header for the CSV\n",
        "empty_df.to_csv(SUBJECT_NOT_FOUND_FILE, header=True, index=False)\n",
        "\n",
        "sample_df[['subject', 'subject_masked_sentence', 'subject_true_label', 'subject_false_label']] = sample_df.apply(generate_subject_data, axis=1, result_type='expand')\n",
        "# drop data that was failed rows \n",
        "sample_df = sample_df.dropna()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LESkkiGr4mtC",
        "outputId": "bfb82c4d-9158-43b3-e295-d27170086039"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "<ipython-input-8-c30e4b5fef12>:1: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  sample_df[['subject_true_label_p', 'subject_false_label_p', 'subject_is_accurate']] = sample_df.apply(lambda x: get_label_probs(x, 'subject'), axis=1, result_type='expand')\n",
            "<ipython-input-8-c30e4b5fef12>:1: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  sample_df[['subject_true_label_p', 'subject_false_label_p', 'subject_is_accurate']] = sample_df.apply(lambda x: get_label_probs(x, 'subject'), axis=1, result_type='expand')\n",
            "<ipython-input-8-c30e4b5fef12>:1: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  sample_df[['subject_true_label_p', 'subject_false_label_p', 'subject_is_accurate']] = sample_df.apply(lambda x: get_label_probs(x, 'subject'), axis=1, result_type='expand')\n"
          ]
        }
      ],
      "source": [
        "sample_df[['subject_true_label_p', 'subject_false_label_p', 'subject_is_accurate']] = sample_df.apply(lambda x: get_label_probs(x, 'subject'), axis=1, result_type='expand')\n",
        "\n",
        "#persist the dataset for easy access of results\n",
        "# NOTE: this file has also been uploaded offline for easier access when compling\n",
        "# results\n",
        "sample_df.to_csv('subject_number_agreement_with_results.csv')\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "id": "deZ9i61Ev24w"
      },
      "outputs": [],
      "source": [
        "# Calculate the results for the nonce dataset\n",
        "nonce_df = pd.read_csv('data/nonce_dataset.csv')\n",
        "\n",
        "nonce_df[['verb_true_label_p', 'verb_false_label_p', 'verb_is_accurate']] = nonce_df.apply(lambda x: get_label_probs(x, 'verb'), axis=1, result_type='expand')\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "id": "DKAR4-JP95zG"
      },
      "outputs": [],
      "source": [
        "nonce_df[['subject', 'subject_masked_sentence', 'subject_true_label', 'subject_false_label']] = nonce_df.apply(generate_subject_data, axis=1, result_type='expand')\n",
        "nonce_df = nonce_df.dropna()\n",
        "nonce_df[['subject_true_label_p', 'subject_false_label_p', 'subject_is_accurate']] = nonce_df.apply(lambda x: get_label_probs(x, 'subject'), axis=1, result_type='expand')\n",
        "\n",
        "nonce_df.to_csv('number_agreement_nonce_with_results.csv')\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WOQ0vnRKo5P2"
      },
      "source": [
        "## <a name=\"large-dataset-results\"></a>Results of the large dataset"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "id": "9J9iiyVH5egL"
      },
      "outputs": [],
      "source": [
        "import locale\n",
        "locale.getpreferredencoding = lambda: \"UTF-8\"\n",
        "\n",
        "# NOTE: You can start running the notebook from here in order to avoid the \n",
        "# slow result processing\n",
        "\n",
        "# download resultfiles if it doesn't exist\n",
        "! [ ! -f subject_number_agreement_with_results.csv ] && wget https://raw.githubusercontent.com/tsido/LDA-T307-project/main/data/subject_number_agreement_with_results.csv"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "id": "a5wtlQQY4mtC"
      },
      "outputs": [],
      "source": [
        "# Reload the results so we don't need to run the model inference every time\n",
        "\n",
        "import pandas as pd\n",
        "result_df = pd.read_csv('subject_number_agreement_with_results.csv')\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 260
        },
        "id": "gTMHbm_eIV-Q",
        "outputId": "40fb3fbe-0cfc-47b0-d2d8-e88036deb03d"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Shared logits for subjects 1817\n",
            "Shared logits for verbs 103\n",
            "Accuracy of BERT with Linzen Goldberg dataset by attractors on unfiltered results:\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "            verb_is_accurate  subject_is_accurate\n",
              "attractors                                       \n",
              "1                   0.955277             0.901920\n",
              "2                   0.948118             0.881772\n",
              "3                   0.941935             0.854839\n",
              "4                   0.928287             0.864542"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-e20f7422-f149-4a1c-b51e-e8a30bb12805\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>verb_is_accurate</th>\n",
              "      <th>subject_is_accurate</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>attractors</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>0.955277</td>\n",
              "      <td>0.901920</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>0.948118</td>\n",
              "      <td>0.881772</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0.941935</td>\n",
              "      <td>0.854839</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>0.928287</td>\n",
              "      <td>0.864542</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-e20f7422-f149-4a1c-b51e-e8a30bb12805')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-e20f7422-f149-4a1c-b51e-e8a30bb12805 button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-e20f7422-f149-4a1c-b51e-e8a30bb12805');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 13
        }
      ],
      "source": [
        "\n",
        "\n",
        "print('Shared logits for subjects', len(result_df.query('subject_true_label_p == subject_false_label_p')))\n",
        "print('Shared logits for verbs', len(result_df.query('verb_true_label_p == verb_false_label_p')))\n",
        "\n",
        "print(\"Accuracy of BERT with Linzen Goldberg dataset by attractors on unfiltered results:\")\n",
        "results = result_df.groupby('attractors')[['verb_is_accurate', 'subject_is_accurate']].mean()\n",
        "results"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 224
        },
        "id": "Jv0vaPyzDObP",
        "outputId": "f65f496f-003f-4d8e-f9e8-c27c42ff7751"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Accuracy of BERT with Linzen Goldberg dataset by attractors with unknown tokens filtered:\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "            verb_is_accurate  subject_is_accurate\n",
              "attractors                                       \n",
              "1                   0.959031             0.960777\n",
              "2                   0.954959             0.940684\n",
              "3                   0.947674             0.923256\n",
              "4                   0.952174             0.934783"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-4a3df885-8989-4067-a612-a9f18aa8227d\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>verb_is_accurate</th>\n",
              "      <th>subject_is_accurate</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>attractors</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>0.959031</td>\n",
              "      <td>0.960777</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>0.954959</td>\n",
              "      <td>0.940684</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0.947674</td>\n",
              "      <td>0.923256</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>0.952174</td>\n",
              "      <td>0.934783</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-4a3df885-8989-4067-a612-a9f18aa8227d')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-4a3df885-8989-4067-a612-a9f18aa8227d button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-4a3df885-8989-4067-a612-a9f18aa8227d');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 14
        }
      ],
      "source": [
        "\n",
        "# Filter the results where the logits for true and false are equal, as those indicate that the token is unknown\n",
        "filtered_results = result_df[(result_df['subject_true_label_p'] != result_df['subject_false_label_p']) & (result_df['verb_true_label_p'] != result_df['verb_false_label_p'])]\n",
        "print(\"Accuracy of BERT with Linzen Goldberg dataset by attractors with unknown tokens filtered:\")\n",
        "accuracy_results = filtered_results.groupby('attractors')[['verb_is_accurate', 'subject_is_accurate']].mean()\n",
        "accuracy_results"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1018
        },
        "id": "K2Yrhe4r4ms_",
        "outputId": "b8ed3409-9781-4e1d-fbd2-7d9a068a57af"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Sentences where both verb and subject is incorrectly chosen by the model:\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Unnamed: 0</th>\n",
              "      <th>attractors</th>\n",
              "      <th>full_sentence</th>\n",
              "      <th>verb_masked_sentence</th>\n",
              "      <th>verb_true_label</th>\n",
              "      <th>verb_false_label</th>\n",
              "      <th>verb_true_label_p</th>\n",
              "      <th>verb_false_label_p</th>\n",
              "      <th>verb_is_accurate</th>\n",
              "      <th>subject</th>\n",
              "      <th>subject_masked_sentence</th>\n",
              "      <th>subject_true_label</th>\n",
              "      <th>subject_false_label</th>\n",
              "      <th>subject_true_label_p</th>\n",
              "      <th>subject_false_label_p</th>\n",
              "      <th>subject_is_accurate</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>122</th>\n",
              "      <td>126</td>\n",
              "      <td>1</td>\n",
              "      <td>according to barsalou , simulation composed of...</td>\n",
              "      <td>according to barsalou , simulation composed of...</td>\n",
              "      <td>creates</td>\n",
              "      <td>create</td>\n",
              "      <td>3.418266</td>\n",
              "      <td>4.212551</td>\n",
              "      <td>False</td>\n",
              "      <td>simulation</td>\n",
              "      <td>according to barsalou , [MASK] composed of the...</td>\n",
              "      <td>simulation</td>\n",
              "      <td>simulations</td>\n",
              "      <td>1.833017</td>\n",
              "      <td>1.904535</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>790</th>\n",
              "      <td>807</td>\n",
              "      <td>2</td>\n",
              "      <td>air force panel ( who were assigned as `` merc...</td>\n",
              "      <td>air force panel ( who were assigned as `` merc...</td>\n",
              "      <td>claims</td>\n",
              "      <td>claim</td>\n",
              "      <td>4.231212</td>\n",
              "      <td>5.418381</td>\n",
              "      <td>False</td>\n",
              "      <td>panel</td>\n",
              "      <td>air force [MASK] ( who were assigned as `` mer...</td>\n",
              "      <td>panel</td>\n",
              "      <td>panels</td>\n",
              "      <td>2.329565</td>\n",
              "      <td>3.139415</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1031</th>\n",
              "      <td>1055</td>\n",
              "      <td>2</td>\n",
              "      <td>all rulings after 1570 as king of hungary refe...</td>\n",
              "      <td>all rulings after 1570 as king of hungary refe...</td>\n",
              "      <td>refer</td>\n",
              "      <td>refers</td>\n",
              "      <td>2.357046</td>\n",
              "      <td>4.353528</td>\n",
              "      <td>False</td>\n",
              "      <td>rulings</td>\n",
              "      <td>all [MASK] after 1570 as king of hungary refer...</td>\n",
              "      <td>rulings</td>\n",
              "      <td>ruling</td>\n",
              "      <td>0.777411</td>\n",
              "      <td>5.986384</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1072</th>\n",
              "      <td>1097</td>\n",
              "      <td>1</td>\n",
              "      <td>all the different fonts and color tend to thro...</td>\n",
              "      <td>all the different fonts and color [MASK] to th...</td>\n",
              "      <td>tend</td>\n",
              "      <td>tends</td>\n",
              "      <td>5.079307</td>\n",
              "      <td>6.851832</td>\n",
              "      <td>False</td>\n",
              "      <td>fonts</td>\n",
              "      <td>all the different [MASK] and color tend to thr...</td>\n",
              "      <td>fonts</td>\n",
              "      <td>font</td>\n",
              "      <td>-0.904954</td>\n",
              "      <td>6.012044</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1450</th>\n",
              "      <td>1484</td>\n",
              "      <td>2</td>\n",
              "      <td>although the record and names of their childre...</td>\n",
              "      <td>although the record and names of their childre...</td>\n",
              "      <td>shows</td>\n",
              "      <td>show</td>\n",
              "      <td>8.341554</td>\n",
              "      <td>8.785891</td>\n",
              "      <td>False</td>\n",
              "      <td>record</td>\n",
              "      <td>although the [MASK] and names of their childre...</td>\n",
              "      <td>record</td>\n",
              "      <td>records</td>\n",
              "      <td>7.086719</td>\n",
              "      <td>9.268765</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>28221</th>\n",
              "      <td>28778</td>\n",
              "      <td>3</td>\n",
              "      <td>update that fixes issues with limited games an...</td>\n",
              "      <td>update that fixes issues with limited games an...</td>\n",
              "      <td>makes</td>\n",
              "      <td>make</td>\n",
              "      <td>13.074266</td>\n",
              "      <td>15.345674</td>\n",
              "      <td>False</td>\n",
              "      <td>fixes</td>\n",
              "      <td>update that [MASK] issues with limited games a...</td>\n",
              "      <td>fixes</td>\n",
              "      <td>fix</td>\n",
              "      <td>0.952856</td>\n",
              "      <td>7.454342</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>28373</th>\n",
              "      <td>28933</td>\n",
              "      <td>1</td>\n",
              "      <td>usually these group of people goes around at n...</td>\n",
              "      <td>usually these group of people [MASK] around at...</td>\n",
              "      <td>goes</td>\n",
              "      <td>go</td>\n",
              "      <td>6.570880</td>\n",
              "      <td>10.423776</td>\n",
              "      <td>False</td>\n",
              "      <td>group</td>\n",
              "      <td>usually these [MASK] of people goes around at ...</td>\n",
              "      <td>group</td>\n",
              "      <td>groups</td>\n",
              "      <td>12.927755</td>\n",
              "      <td>13.057397</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>28668</th>\n",
              "      <td>29236</td>\n",
              "      <td>1</td>\n",
              "      <td>weak keep it is referred to a lot by various w...</td>\n",
              "      <td>weak keep it is referred to a lot by various w...</td>\n",
              "      <td>gets</td>\n",
              "      <td>get</td>\n",
              "      <td>4.836196</td>\n",
              "      <td>6.246789</td>\n",
              "      <td>False</td>\n",
              "      <td>lifestyle</td>\n",
              "      <td>weak keep it is referred to a lot by various w...</td>\n",
              "      <td>lifestyle</td>\n",
              "      <td>lifestyles</td>\n",
              "      <td>0.758718</td>\n",
              "      <td>0.990094</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>28671</th>\n",
              "      <td>29239</td>\n",
              "      <td>1</td>\n",
              "      <td>weak oppose slanted posts , glare on the hat ,...</td>\n",
              "      <td>weak oppose slanted posts , glare on the hat ,...</td>\n",
              "      <td>looks</td>\n",
              "      <td>look</td>\n",
              "      <td>3.129375</td>\n",
              "      <td>3.430743</td>\n",
              "      <td>False</td>\n",
              "      <td>weak</td>\n",
              "      <td>[MASK] oppose slanted posts , glare on the hat...</td>\n",
              "      <td>weak</td>\n",
              "      <td>weaks</td>\n",
              "      <td>-1.480049</td>\n",
              "      <td>2.497913</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>28937</th>\n",
              "      <td>29507</td>\n",
              "      <td>1</td>\n",
              "      <td>while in skinner 's office , the camera angles...</td>\n",
              "      <td>while in skinner 's office , the camera angles...</td>\n",
              "      <td>reveals</td>\n",
              "      <td>reveal</td>\n",
              "      <td>10.740785</td>\n",
              "      <td>12.171120</td>\n",
              "      <td>False</td>\n",
              "      <td>angles</td>\n",
              "      <td>while in skinner 's office , the camera [MASK]...</td>\n",
              "      <td>angles</td>\n",
              "      <td>angle</td>\n",
              "      <td>11.892818</td>\n",
              "      <td>13.314294</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>93 rows × 16 columns</p>\n",
              "</div>"
            ],
            "application/vnd.google.colaboratory.module+javascript": "\n      import \"https://ssl.gstatic.com/colaboratory/data_table/99dac6621f6ae8c4/data_table.js\";\n\n      window.createDataTable({\n        data: [[{\n            'v': 126,\n            'f': \"126\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"according to barsalou , simulation composed of these three modalities creates representations and all of the contents of cognition .\",\n\"according to barsalou , simulation composed of these three modalities [MASK] representations and all of the contents of cognition .\",\n\"creates\",\n\"create\",\n{\n            'v': 3.4182662963867188,\n            'f': \"3.4182662963867188\",\n        },\n{\n            'v': 4.212551116943359,\n            'f': \"4.212551116943359\",\n        },\nfalse,\n\"simulation\",\n\"according to barsalou , [MASK] composed of these three modalities creates representations and all of the contents of cognition .\",\n\"simulation\",\n\"simulations\",\n{\n            'v': 1.8330165147781368,\n            'f': \"1.8330165147781368\",\n        },\n{\n            'v': 1.9045348167419436,\n            'f': \"1.9045348167419436\",\n        },\nfalse],\n [{\n            'v': 807,\n            'f': \"807\",\n        },\n{\n            'v': 2,\n            'f': \"2\",\n        },\n\"air force panel ( who were assigned as `` merchant marine experts '' on the veterans claims board ) lost , he still refused to grant veteran 's status .\",\n\"air force panel ( who were assigned as `` merchant marine experts '' on the veterans [MASK] board ) lost , he still refused to grant veteran 's status .\",\n\"claims\",\n\"claim\",\n{\n            'v': 4.231212139129639,\n            'f': \"4.231212139129639\",\n        },\n{\n            'v': 5.418381214141846,\n            'f': \"5.418381214141846\",\n        },\nfalse,\n\"panel\",\n\"air force [MASK] ( who were assigned as `` merchant marine experts '' on the veterans claims board ) lost , he still refused to grant veteran 's status .\",\n\"panel\",\n\"panels\",\n{\n            'v': 2.3295652866363525,\n            'f': \"2.3295652866363525\",\n        },\n{\n            'v': 3.1394147872924805,\n            'f': \"3.1394147872924805\",\n        },\nfalse],\n [{\n            'v': 1055,\n            'f': \"1055\",\n        },\n{\n            'v': 2,\n            'f': \"2\",\n        },\n\"all rulings after 1570 as king of hungary refer to the territory known as `` royal hungary '' , and as prince refer to the `` principality of transylvania '' .\",\n\"all rulings after 1570 as king of hungary refer to the territory known as `` royal hungary '' , and as prince [MASK] to the `` principality of transylvania '' .\",\n\"refer\",\n\"refers\",\n{\n            'v': 2.357045888900757,\n            'f': \"2.357045888900757\",\n        },\n{\n            'v': 4.353528499603272,\n            'f': \"4.353528499603272\",\n        },\nfalse,\n\"rulings\",\n\"all [MASK] after 1570 as king of hungary refer to the territory known as `` royal hungary '' , and as prince refer to the `` principality of transylvania '' .\",\n\"rulings\",\n\"ruling\",\n{\n            'v': 0.7774105668067932,\n            'f': \"0.7774105668067932\",\n        },\n{\n            'v': 5.986384391784668,\n            'f': \"5.986384391784668\",\n        },\nfalse],\n [{\n            'v': 1097,\n            'f': \"1097\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"all the different fonts and color tend to throw me off .\",\n\"all the different fonts and color [MASK] to throw me off .\",\n\"tend\",\n\"tends\",\n{\n            'v': 5.079306602478027,\n            'f': \"5.079306602478027\",\n        },\n{\n            'v': 6.851831912994385,\n            'f': \"6.851831912994385\",\n        },\nfalse,\n\"fonts\",\n\"all the different [MASK] and color tend to throw me off .\",\n\"fonts\",\n\"font\",\n{\n            'v': -0.904954433441162,\n            'f': \"-0.904954433441162\",\n        },\n{\n            'v': 6.012044429779053,\n            'f': \"6.012044429779053\",\n        },\nfalse],\n [{\n            'v': 1484,\n            'f': \"1484\",\n        },\n{\n            'v': 2,\n            'f': \"2\",\n        },\n\"although the record and names of their children , taken from the will of stephen durnan dated july 7 , 1888 shows the 12 children listed , at the gravesite in immaculate conception cemetery , fairbank , iowa .\",\n\"although the record and names of their children , taken from the will of stephen durnan dated july 7 , 1888 [MASK] the 12 children listed , at the gravesite in immaculate conception cemetery , fairbank , iowa .\",\n\"shows\",\n\"show\",\n{\n            'v': 8.341553688049316,\n            'f': \"8.341553688049316\",\n        },\n{\n            'v': 8.785890579223633,\n            'f': \"8.785890579223633\",\n        },\nfalse,\n\"record\",\n\"although the [MASK] and names of their children , taken from the will of stephen durnan dated july 7 , 1888 shows the 12 children listed , at the gravesite in immaculate conception cemetery , fairbank , iowa .\",\n\"record\",\n\"records\",\n{\n            'v': 7.086718559265137,\n            'f': \"7.086718559265137\",\n        },\n{\n            'v': 9.268765449523926,\n            'f': \"9.268765449523926\",\n        },\nfalse],\n [{\n            'v': 1893,\n            'f': \"1893\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"an orchard of 900 trees grows apples , pears and cherries .\",\n\"an orchard of 900 trees [MASK] apples , pears and cherries .\",\n\"grows\",\n\"grow\",\n{\n            'v': 7.701584339141846,\n            'f': \"7.701584339141846\",\n        },\n{\n            'v': 8.101618766784668,\n            'f': \"8.101618766784668\",\n        },\nfalse,\n\"orchard\",\n\"an [MASK] of 900 trees grows apples , pears and cherries .\",\n\"orchard\",\n\"orchards\",\n{\n            'v': 3.94409441947937,\n            'f': \"3.94409441947937\",\n        },\n{\n            'v': 4.377961158752441,\n            'f': \"4.377961158752441\",\n        },\nfalse],\n [{\n            'v': 1947,\n            'f': \"1947\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"another , the value of which plays a huge role in determining the nature of our universe , is the gravitational coupling constant , the ratio of gravitational to electromagnetic forces expresses as a dimensionless value -- it is about 1 .\",\n\"another , the value of which plays a huge role in determining the nature of our universe , is the gravitational coupling constant , the ratio of gravitational to electromagnetic forces [MASK] as a dimensionless value -- it is about 1 .\",\n\"expresses\",\n\"express\",\n{\n            'v': 3.362454891204834,\n            'f': \"3.362454891204834\",\n        },\n{\n            'v': 3.604464530944824,\n            'f': \"3.604464530944824\",\n        },\nfalse,\n\"another\",\n\"[MASK] , the value of which plays a huge role in determining the nature of our universe , is the gravitational coupling constant , the ratio of gravitational to electromagnetic forces expresses as a dimensionless value -- it is about 1 .\",\n\"another\",\n\"anothers\",\n{\n            'v': 4.619450092315674,\n            'f': \"4.619450092315674\",\n        },\n{\n            'v': 4.96886157989502,\n            'f': \"4.96886157989502\",\n        },\nfalse],\n [{\n            'v': 2964,\n            'f': \"2964\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"at least one hang gliding businesses operates from bald hill , which is technically a part of stanwell tops .\",\n\"at least one hang gliding businesses [MASK] from bald hill , which is technically a part of stanwell tops .\",\n\"operates\",\n\"operate\",\n{\n            'v': 10.875499725341797,\n            'f': \"10.875499725341797\",\n        },\n{\n            'v': 13.132437705993652,\n            'f': \"13.132437705993652\",\n        },\nfalse,\n\"businesses\",\n\"at least one hang gliding [MASK] operates from bald hill , which is technically a part of stanwell tops .\",\n\"businesses\",\n\"business\",\n{\n            'v': -1.181623935699463,\n            'f': \"-1.181623935699463\",\n        },\n{\n            'v': 5.289834976196289,\n            'f': \"5.289834976196289\",\n        },\nfalse],\n [{\n            'v': 3096,\n            'f': \"3096\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"australian data sheet for patients adds .\",\n\"australian data sheet for patients [MASK] .\",\n\"adds\",\n\"add\",\n{\n            'v': -2.0285985469818115,\n            'f': \"-2.0285985469818115\",\n        },\n{\n            'v': 0.2432263791561126,\n            'f': \"0.2432263791561126\",\n        },\nfalse,\n\"sheet\",\n\"australian data [MASK] for patients adds .\",\n\"sheet\",\n\"sheets\",\n{\n            'v': 5.076869964599609,\n            'f': \"5.076869964599609\",\n        },\n{\n            'v': 5.109294891357422,\n            'f': \"5.109294891357422\",\n        },\nfalse],\n [{\n            'v': 3221,\n            'f': \"3221\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"bart 's fantasy where he sees the founding fathers of the united states uses muted colors and variations of red , white and blue .\",\n\"bart 's fantasy where he sees the founding fathers of the united states [MASK] muted colors and variations of red , white and blue .\",\n\"uses\",\n\"use\",\n{\n            'v': 3.8464019298553462,\n            'f': \"3.8464019298553462\",\n        },\n{\n            'v': 6.525009632110596,\n            'f': \"6.525009632110596\",\n        },\nfalse,\n\"he\",\n\"bart 's fantasy w[MASK]re [MASK] sees t[MASK] founding fat[MASK]rs of t[MASK] united states uses muted colors and variations of red , white and blue .\",\n\"he\",\n\"they\",\n{\n            'v': 4.200623512268066,\n            'f': \"4.200623512268066\",\n        },\n{\n            'v': 4.232672691345215,\n            'f': \"4.232672691345215\",\n        },\nfalse],\n [{\n            'v': 3247,\n            'f': \"3247\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"battles after hidalgo 's deathon march 26 , 1811 , ray\\u00f3n move out from saltillo to zacatecas in that moment his army was composed by 3500 soldiers and 22 guns .\",\n\"battles after hidalgo 's deathon march 26 , 1811 , ray\\u00f3n [MASK] out from saltillo to zacatecas in that moment his army was composed by 3500 soldiers and 22 guns .\",\n\"move\",\n\"moves\",\n{\n            'v': 4.914750099182129,\n            'f': \"4.914750099182129\",\n        },\n{\n            'v': 6.333975315093994,\n            'f': \"6.333975315093994\",\n        },\nfalse,\n\"battles\",\n\"[MASK] after hidalgo 's deathon march 26 , 1811 , ray\\u00f3n move out from saltillo to zacatecas in that moment his army was composed by 3500 soldiers and 22 guns .\",\n\"battles\",\n\"battle\",\n{\n            'v': -1.5456911325454712,\n            'f': \"-1.5456911325454712\",\n        },\n{\n            'v': -0.1261520385742187,\n            'f': \"-0.1261520385742187\",\n        },\nfalse],\n [{\n            'v': 3667,\n            'f': \"3667\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"burroughs concept of a dying mars and the martian canals follows the theories of lowell and his predecessor giovanni schiaparelli .\",\n\"burroughs concept of a dying mars and the martian canals [MASK] the theories of lowell and his predecessor giovanni schiaparelli .\",\n\"follows\",\n\"follow\",\n{\n            'v': 6.360954284667969,\n            'f': \"6.360954284667969\",\n        },\n{\n            'v': 7.531383991241455,\n            'f': \"7.531383991241455\",\n        },\nfalse,\n\"canals\",\n\"burroughs concept of a dying mars and the martian [MASK] follows the theories of lowell and his predecessor giovanni schiaparelli .\",\n\"canals\",\n\"canal\",\n{\n            'v': 0.1443674266338348,\n            'f': \"0.1443674266338348\",\n        },\n{\n            'v': 1.3149408102035522,\n            'f': \"1.3149408102035522\",\n        },\nfalse],\n [{\n            'v': 3770,\n            'f': \"3770\",\n        },\n{\n            'v': 3,\n            'f': \"3\",\n        },\n\"but the relatively high levels of presentiation and discreteness in the advanced technology of post-technique require a good bit of law implementing these norms .\",\n\"but the relatively high levels of presentiation and discreteness in the advanced technology of post-technique [MASK] a good bit of law implementing these norms .\",\n\"require\",\n\"requires\",\n{\n            'v': 8.457892417907715,\n            'f': \"8.457892417907715\",\n        },\n{\n            'v': 9.93846035003662,\n            'f': \"9.93846035003662\",\n        },\nfalse,\n\"levels\",\n\"but the relatively high [MASK] of presentiation and discreteness in the advanced technology of post-technique require a good bit of law implementing these norms .\",\n\"levels\",\n\"level\",\n{\n            'v': 12.574481964111328,\n            'f': \"12.574481964111328\",\n        },\n{\n            'v': 12.577247619628906,\n            'f': \"12.577247619628906\",\n        },\nfalse],\n [{\n            'v': 3794,\n            'f': \"3794\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"by 2010 , the regreening programs had successfully rehabilitated 3 , 350 hectares of land in the city ; however , approximately 30 , 000 hectares of land have yet to be rehabilitated .\",\n\"by 2010 , the regreening programs had successfully rehabilitated 3 , 350 hectares of land in the city ; however , approximately 30 , 000 hectares of land [MASK] yet to be rehabilitated .\",\n\"have\",\n\"has\",\n{\n            'v': 15.675664901733398,\n            'f': \"15.675664901733398\",\n        },\n{\n            'v': 16.045381546020508,\n            'f': \"16.045381546020508\",\n        },\nfalse,\n\"programs\",\n\"by 2010 , the regreening [MASK] had successfully rehabilitated 3 , 350 hectares of land in the city ; however , approximately 30 , 000 hectares of land have yet to be rehabilitated .\",\n\"programs\",\n\"program\",\n{\n            'v': 7.475152015686035,\n            'f': \"7.475152015686035\",\n        },\n{\n            'v': 11.26876163482666,\n            'f': \"11.26876163482666\",\n        },\nfalse],\n [{\n            'v': 3982,\n            'f': \"3982\",\n        },\n{\n            'v': 3,\n            'f': \"3\",\n        },\n\"centrifugal forces and the fact that the object is moving in its own wake mean that detailed examination of the airflow is difficult .\",\n\"centrifugal forces and the fact that the object is moving in its own wake [MASK] that detailed examination of the airflow is difficult .\",\n\"mean\",\n\"means\",\n{\n            'v': 14.355990409851074,\n            'f': \"14.355990409851074\",\n        },\n{\n            'v': 14.676008224487305,\n            'f': \"14.676008224487305\",\n        },\nfalse,\n\"forces\",\n\"centrifugal [MASK] and the fact that the object is moving in its own wake mean that detailed examination of the airflow is difficult .\",\n\"forces\",\n\"force\",\n{\n            'v': 15.900251388549805,\n            'f': \"15.900251388549805\",\n        },\n{\n            'v': 16.722482681274414,\n            'f': \"16.722482681274414\",\n        },\nfalse],\n [{\n            'v': 4137,\n            'f': \"4137\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"choose how the search box buttons and frame appear .\",\n\"choose how the search box buttons and frame [MASK] .\",\n\"appear\",\n\"appears\",\n{\n            'v': 6.613608360290527,\n            'f': \"6.613608360290527\",\n        },\n{\n            'v': 7.838123321533203,\n            'f': \"7.838123321533203\",\n        },\nfalse,\n\"buttons\",\n\"choose how the search box [MASK] and frame appear .\",\n\"buttons\",\n\"button\",\n{\n            'v': 4.061965465545654,\n            'f': \"4.061965465545654\",\n        },\n{\n            'v': 4.803630352020264,\n            'f': \"4.803630352020264\",\n        },\nfalse],\n [{\n            'v': 4717,\n            'f': \"4717\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"cooking technique , and a new process in regards to pasta , follows in this article .\",\n\"cooking technique , and a new process in regards to pasta , [MASK] in this article .\",\n\"follows\",\n\"follow\",\n{\n            'v': 4.897762298583984,\n            'f': \"4.897762298583984\",\n        },\n{\n            'v': 6.202620983123779,\n            'f': \"6.202620983123779\",\n        },\nfalse,\n\"technique\",\n\"cooking [MASK] , and a new process in regards to pasta , follows in this article .\",\n\"technique\",\n\"techniques\",\n{\n            'v': 7.925588130950928,\n            'f': \"7.925588130950928\",\n        },\n{\n            'v': 9.646249771118164,\n            'f': \"9.646249771118164\",\n        },\nfalse],\n [{\n            'v': 4978,\n            'f': \"4978\",\n        },\n{\n            'v': 4,\n            'f': \"4\",\n        },\n\"dahl states that difference in size , diversity , and economic affluence make attributing good performance on these measures to government effectiveness too difficult .\",\n\"dahl states that difference in size , diversity , and economic affluence [MASK] attributing good performance on these measures to government effectiveness too difficult .\",\n\"make\",\n\"makes\",\n{\n            'v': 19.9039306640625,\n            'f': \"19.9039306640625\",\n        },\n{\n            'v': 20.103673934936523,\n            'f': \"20.103673934936523\",\n        },\nfalse,\n\"difference\",\n\"dahl states that [MASK] in size , diversity , and economic affluence make attributing good performance on these measures to government effectiveness too difficult .\",\n\"difference\",\n\"differences\",\n{\n            'v': 9.785890579223633,\n            'f': \"9.785890579223633\",\n        },\n{\n            'v': 15.77769947052002,\n            'f': \"15.77769947052002\",\n        },\nfalse],\n [{\n            'v': 5400,\n            'f': \"5400\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"don honors her request , but neither of them get any sleep that night .\",\n\"don honors her request , but neither of them [MASK] any sleep that night .\",\n\"get\",\n\"gets\",\n{\n            'v': 14.548717498779297,\n            'f': \"14.548717498779297\",\n        },\n{\n            'v': 16.466283798217773,\n            'f': \"16.466283798217773\",\n        },\nfalse,\n\"don\",\n\"[MASK] honors her request , but neither of them get any sleep that night .\",\n\"don\",\n\"dons\",\n{\n            'v': -1.9147971868515012,\n            'f': \"-1.9147971868515012\",\n        },\n{\n            'v': 0.0833533704280853,\n            'f': \"0.0833533704280853\",\n        },\nfalse],\n [{\n            'v': 5892,\n            'f': \"5892\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"even relatively poor countries as czech republic have something .\",\n\"even relatively poor countries as czech republic [MASK] something .\",\n\"have\",\n\"has\",\n{\n            'v': 3.9178578853607178,\n            'f': \"3.9178578853607178\",\n        },\n{\n            'v': 4.4553046226501465,\n            'f': \"4.4553046226501465\",\n        },\nfalse,\n\"republic\",\n\"even relatively poor countries as czech [MASK] have something .\",\n\"republic\",\n\"republics\",\n{\n            'v': -1.6436558961868286,\n            'f': \"-1.6436558961868286\",\n        },\n{\n            'v': 2.7740910053253174,\n            'f': \"2.7740910053253174\",\n        },\nfalse],\n [{\n            'v': 6476,\n            'f': \"6476\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"field hockey players by century , merge the rest .\",\n\"field hockey players by century , [MASK] the rest .\",\n\"merge\",\n\"merges\",\n{\n            'v': -0.3159836232662201,\n            'f': \"-0.3159836232662201\",\n        },\n{\n            'v': 1.3059428930282593,\n            'f': \"1.3059428930282593\",\n        },\nfalse,\n\"players\",\n\"field hockey [MASK] by century , merge the rest .\",\n\"players\",\n\"player\",\n{\n            'v': 1.5399630069732666,\n            'f': \"1.5399630069732666\",\n        },\n{\n            'v': 1.7006274461746216,\n            'f': \"1.7006274461746216\",\n        },\nfalse],\n [{\n            'v': 6525,\n            'f': \"6525\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"finally , the emergency powers passed by the althing on the night of 6 october contain a measure ( art .\",\n\"finally , the emergency powers passed by the althing on the night of 6 october [MASK] a measure ( art .\",\n\"contain\",\n\"contains\",\n{\n            'v': 0.0752125978469848,\n            'f': \"0.0752125978469848\",\n        },\n{\n            'v': 1.3525283336639404,\n            'f': \"1.3525283336639404\",\n        },\nfalse,\n\"powers\",\n\"finally , the emergency [MASK] passed by the althing on the night of 6 october contain a measure ( art .\",\n\"powers\",\n\"power\",\n{\n            'v': 4.009816646575928,\n            'f': \"4.009816646575928\",\n        },\n{\n            'v': 4.897215843200684,\n            'f': \"4.897215843200684\",\n        },\nfalse],\n [{\n            'v': 6742,\n            'f': \"6742\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"for example , the thai currency , stamps , etc still uses images that are probably 40-50 years old .\",\n\"for example , the thai currency , stamps , etc still [MASK] images that are probably 40-50 years old .\",\n\"uses\",\n\"use\",\n{\n            'v': 4.948536396026611,\n            'f': \"4.948536396026611\",\n        },\n{\n            'v': 9.551706314086914,\n            'f': \"9.551706314086914\",\n        },\nfalse,\n\"currency\",\n\"for example , the thai [MASK] , stamps , etc still uses images that are probably 40-50 years old .\",\n\"currency\",\n\"currencies\",\n{\n            'v': 5.718693733215332,\n            'f': \"5.718693733215332\",\n        },\n{\n            'v': 5.991974830627441,\n            'f': \"5.991974830627441\",\n        },\nfalse],\n [{\n            'v': 7182,\n            'f': \"7182\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"geometric period of vases ends in ancient greece .\",\n\"geometric period of vases [MASK] in ancient greece .\",\n\"ends\",\n\"end\",\n{\n            'v': 2.321041822433472,\n            'f': \"2.321041822433472\",\n        },\n{\n            'v': 2.6353752613067627,\n            'f': \"2.6353752613067627\",\n        },\nfalse,\n\"period\",\n\"geometric [MASK] of vases ends in ancient greece .\",\n\"period\",\n\"periods\",\n{\n            'v': 2.126093864440918,\n            'f': \"2.126093864440918\",\n        },\n{\n            'v': 2.39450740814209,\n            'f': \"2.39450740814209\",\n        },\nfalse],\n [{\n            'v': 7250,\n            'f': \"7250\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"gold , silver , cooper and iron could be easily found but the remote mountainous area and the lack of roads makes non-profitable their exploitation .\",\n\"gold , silver , cooper and iron could be easily found but the remote mountainous area and the lack of roads [MASK] non-profitable their exploitation .\",\n\"makes\",\n\"make\",\n{\n            'v': 9.224613189697266,\n            'f': \"9.224613189697266\",\n        },\n{\n            'v': 10.03290843963623,\n            'f': \"10.03290843963623\",\n        },\nfalse,\n\"area\",\n\"gold , silver , cooper and iron could be easily found but the remote mountainous [MASK] and the lack of roads makes non-profitable their exploitation .\",\n\"area\",\n\"areas\",\n{\n            'v': 11.827770233154297,\n            'f': \"11.827770233154297\",\n        },\n{\n            'v': 12.61861515045166,\n            'f': \"12.61861515045166\",\n        },\nfalse],\n [{\n            'v': 7280,\n            'f': \"7280\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"google scholar and books brings up only items written by this author , no citations or mentions by other others so there does n't appear to be any critical review of his work .\",\n\"google scholar and books [MASK] up only items written by this author , no citations or mentions by other others so there does n't appear to be any critical review of his work .\",\n\"brings\",\n\"bring\",\n{\n            'v': 8.569028854370117,\n            'f': \"8.569028854370117\",\n        },\n{\n            'v': 11.692959785461426,\n            'f': \"11.692959785461426\",\n        },\nfalse,\n\"scholar\",\n\"google [MASK] and books brings up only items written by this author , no citations or mentions by other others so there does n't appear to be any critical review of his work .\",\n\"scholar\",\n\"scholars\",\n{\n            'v': 1.096936583518982,\n            'f': \"1.096936583518982\",\n        },\n{\n            'v': 3.802032470703125,\n            'f': \"3.802032470703125\",\n        },\nfalse],\n [{\n            'v': 7281,\n            'f': \"7281\",\n        },\n{\n            'v': 3,\n            'f': \"3\",\n        },\n\"google scholar 's only hits on `` hydrogen fuel injection '' seem to involve scramjets .\",\n\"google scholar 's only hits on `` hydrogen fuel injection '' [MASK] to involve scramjets .\",\n\"seem\",\n\"seems\",\n{\n            'v': 9.987772941589355,\n            'f': \"9.987772941589355\",\n        },\n{\n            'v': 11.015487670898438,\n            'f': \"11.015487670898438\",\n        },\nfalse,\n\"hits\",\n\"google scholar 's only [MASK] on `` hydrogen fuel injection '' seem to involve scramjets .\",\n\"hits\",\n\"hit\",\n{\n            'v': 3.6916112899780273,\n            'f': \"3.6916112899780273\",\n        },\n{\n            'v': 3.817363977432251,\n            'f': \"3.817363977432251\",\n        },\nfalse],\n [{\n            'v': 7283,\n            'f': \"7283\",\n        },\n{\n            'v': 2,\n            'f': \"2\",\n        },\n\"google searches for the title of the comic and the author produce no meaningful hits .\",\n\"google searches for the title of the comic and the author [MASK] no meaningful hits .\",\n\"produce\",\n\"produces\",\n{\n            'v': 5.14277982711792,\n            'f': \"5.14277982711792\",\n        },\n{\n            'v': 7.654523849487305,\n            'f': \"7.654523849487305\",\n        },\nfalse,\n\"author\",\n\"google searches for the title of the comic and the [MASK] produce no meaningful hits .\",\n\"author\",\n\"authors\",\n{\n            'v': 4.262138843536377,\n            'f': \"4.262138843536377\",\n        },\n{\n            'v': 7.175344467163086,\n            'f': \"7.175344467163086\",\n        },\nfalse],\n [{\n            'v': 7528,\n            'f': \"7528\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"he lacked the words , so danced instead ; the position of the hands resembles the head and antlers of a stag .\",\n\"he lacked the words , so danced instead ; the position of the hands [MASK] the head and antlers of a stag .\",\n\"resembles\",\n\"resemble\",\n{\n            'v': 2.512732982635498,\n            'f': \"2.512732982635498\",\n        },\n{\n            'v': 2.897149085998535,\n            'f': \"2.897149085998535\",\n        },\nfalse,\n\"he\",\n\"[MASK] lacked t[MASK] words , so danced instead ; t[MASK] position of t[MASK] hands resembles t[MASK] [MASK]ad and antlers of a stag .\",\n\"he\",\n\"they\",\n{\n            'v': 7.989871025085449,\n            'f': \"7.989871025085449\",\n        },\n{\n            'v': 8.024063110351562,\n            'f': \"8.024063110351562\",\n        },\nfalse],\n [{\n            'v': 7959,\n            'f': \"7959\",\n        },\n{\n            'v': 2,\n            'f': \"2\",\n        },\n\"his leg and feet bones makes it clear that neanderthals walked upright like modern humans .\",\n\"his leg and feet bones [MASK] it clear that neanderthals walked upright like modern humans .\",\n\"makes\",\n\"make\",\n{\n            'v': 12.515607833862305,\n            'f': \"12.515607833862305\",\n        },\n{\n            'v': 15.152901649475098,\n            'f': \"15.152901649475098\",\n        },\nfalse,\n\"bones\",\n\"his leg and feet [MASK] makes it clear that neanderthals walked upright like modern humans .\",\n\"bones\",\n\"bone\",\n{\n            'v': 2.326103925704956,\n            'f': \"2.326103925704956\",\n        },\n{\n            'v': 3.2579376697540283,\n            'f': \"3.2579376697540283\",\n        },\nfalse],\n [{\n            'v': 8049,\n            'f': \"8049\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"his responses and editing history show an individual who takes a very reasoned and mature approach to wikipedia .\",\n\"his responses and editing history [MASK] an individual who takes a very reasoned and mature approach to wikipedia .\",\n\"show\",\n\"shows\",\n{\n            'v': 10.630326271057127,\n            'f': \"10.630326271057127\",\n        },\n{\n            'v': 10.96880054473877,\n            'f': \"10.96880054473877\",\n        },\nfalse,\n\"responses\",\n\"his [MASK] and editing history show an individual who takes a very reasoned and mature approach to wikipedia .\",\n\"responses\",\n\"response\",\n{\n            'v': 4.000743389129639,\n            'f': \"4.000743389129639\",\n        },\n{\n            'v': 4.370695114135742,\n            'f': \"4.370695114135742\",\n        },\nfalse],\n [{\n            'v': 8779,\n            'f': \"8779\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"if no editors other than the author show any interest in the article by the time the afd comes to an end , then it can be deleted under this criterion .\",\n\"if no editors other than the author [MASK] any interest in the article by the time the afd comes to an end , then it can be deleted under this criterion .\",\n\"show\",\n\"shows\",\n{\n            'v': 12.05644702911377,\n            'f': \"12.05644702911377\",\n        },\n{\n            'v': 12.23324489593506,\n            'f': \"12.23324489593506\",\n        },\nfalse,\n\"editors\",\n\"if no [MASK] other than the author show any interest in the article by the time the afd comes to an end , then it can be deleted under this criterion .\",\n\"editors\",\n\"editor\",\n{\n            'v': 8.758997917175293,\n            'f': \"8.758997917175293\",\n        },\n{\n            'v': 9.221373558044434,\n            'f': \"9.221373558044434\",\n        },\nfalse],\n [{\n            'v': 10258,\n            'f': \"10258\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"i to air personal grievances only serves to worsen the unhealthy , punitive atmosphere .\",\n\"i to air personal grievances only [MASK] to worsen the unhealthy , punitive atmosphere .\",\n\"serves\",\n\"serve\",\n{\n            'v': 6.182624816894531,\n            'f': \"6.182624816894531\",\n        },\n{\n            'v': 7.332755088806152,\n            'f': \"7.332755088806152\",\n        },\nfalse,\n\"i\",\n\"[MASK] to a[MASK]r personal gr[MASK]evances only serves to worsen the unhealthy , pun[MASK]t[MASK]ve atmosphere .\",\n\"i\",\n\"is\",\n{\n            'v': 0.3389680385589599,\n            'f': \"0.3389680385589599\",\n        },\n{\n            'v': 0.3978262543678283,\n            'f': \"0.3978262543678283\",\n        },\nfalse],\n [{\n            'v': 10360,\n            'f': \"10360\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"its not a problem with the ps3 , because my friends looks the same and changing the font size or any other options does nothing .\",\n\"its not a problem with the ps3 , because my friends [MASK] the same and changing the font size or any other options does nothing .\",\n\"looks\",\n\"look\",\n{\n            'v': 3.991410493850708,\n            'f': \"3.991410493850708\",\n        },\n{\n            'v': 10.073686599731444,\n            'f': \"10.073686599731444\",\n        },\nfalse,\n\"friends\",\n\"its not a problem with the ps3 , because my [MASK] looks the same and changing the font size or any other options does nothing .\",\n\"friends\",\n\"friend\",\n{\n            'v': 0.6538509726524353,\n            'f': \"0.6538509726524353\",\n        },\n{\n            'v': 4.124213218688965,\n            'f': \"4.124213218688965\",\n        },\nfalse],\n [{\n            'v': 10491,\n            'f': \"10491\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"jared diamond in his award-winning books , guns , germs and steel and the third chimpanzee provides sociological and anthropological evidence for the rise of large-scale warfare as a result of advances in technology and city-states .\",\n\"jared diamond in his award-winning books , guns , germs and steel and the third chimpanzee [MASK] sociological and anthropological evidence for the rise of large-scale warfare as a result of advances in technology and city-states .\",\n\"provides\",\n\"provide\",\n{\n            'v': 11.694292068481444,\n            'f': \"11.694292068481444\",\n        },\n{\n            'v': 11.74001121520996,\n            'f': \"11.74001121520996\",\n        },\nfalse,\n\"diamond\",\n\"jared [MASK] in his award-winning books , guns , germs and steel and the third chimpanzee provides sociological and anthropological evidence for the rise of large-scale warfare as a result of advances in technology and city-states .\",\n\"diamond\",\n\"diamonds\",\n{\n            'v': -1.323176622390747,\n            'f': \"-1.323176622390747\",\n        },\n{\n            'v': -0.816946268081665,\n            'f': \"-0.816946268081665\",\n        },\nfalse],\n [{\n            'v': 10802,\n            'f': \"10802\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"last three paragraphs of `` organization of competitions '' has no references .\",\n\"last three paragraphs of `` organization of competitions '' [MASK] no references .\",\n\"has\",\n\"have\",\n{\n            'v': 8.153973579406738,\n            'f': \"8.153973579406738\",\n        },\n{\n            'v': 9.0577392578125,\n            'f': \"9.0577392578125\",\n        },\nfalse,\n\"paragraphs\",\n\"last three [MASK] of `` organization of competitions '' has no references .\",\n\"paragraphs\",\n\"paragraph\",\n{\n            'v': 1.7332913875579834,\n            'f': \"1.7332913875579834\",\n        },\n{\n            'v': 6.695255279541016,\n            'f': \"6.695255279541016\",\n        },\nfalse],\n [{\n            'v': 11017,\n            'f': \"11017\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"list of fictional jews exists , as does list of asian jews , list of lgbt jews and list of lgbt people .\",\n\"list of fictional jews exists , as [MASK] list of asian jews , list of lgbt jews and list of lgbt people .\",\n\"does\",\n\"do\",\n{\n            'v': 9.171249389648438,\n            'f': \"9.171249389648438\",\n        },\n{\n            'v': 9.664764404296877,\n            'f': \"9.664764404296877\",\n        },\nfalse,\n\"list\",\n\"[MASK] of fictional jews exists , as does [MASK] of asian jews , [MASK] of lgbt jews and [MASK] of lgbt people .\",\n\"list\",\n\"lists\",\n{\n            'v': 3.56156325340271,\n            'f': \"3.56156325340271\",\n        },\n{\n            'v': 3.994641304016113,\n            'f': \"3.994641304016113\",\n        },\nfalse],\n [{\n            'v': 11169,\n            'f': \"11169\",\n        },\n{\n            'v': 3,\n            'f': \"3\",\n        },\n\"lot of dovotees from neibhouring places and districts visits the church and holy remains of the chaplain .\",\n\"lot of dovotees from neibhouring places and districts [MASK] the church and holy remains of the chaplain .\",\n\"visits\",\n\"visit\",\n{\n            'v': 5.099145889282227,\n            'f': \"5.099145889282227\",\n        },\n{\n            'v': 7.613540172576904,\n            'f': \"7.613540172576904\",\n        },\nfalse,\n\"lot\",\n\"[MASK] of dovotees from neibhouring places and districts visits the church and holy remains of the chaplain .\",\n\"lot\",\n\"lots\",\n{\n            'v': 1.275403380393982,\n            'f': \"1.275403380393982\",\n        },\n{\n            'v': 5.254663944244385,\n            'f': \"5.254663944244385\",\n        },\nfalse],\n [{\n            'v': 11170,\n            'f': \"11170\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"lot of tourists loves to go for this tremendous fishing before their bbq meal , while the ukulhas island wills to provide fishing gears and all the required facilities on request .\",\n\"lot of tourists [MASK] to go for this tremendous fishing before their bbq meal , while the ukulhas island wills to provide fishing gears and all the required facilities on request .\",\n\"loves\",\n\"love\",\n{\n            'v': 5.745396137237549,\n            'f': \"5.745396137237549\",\n        },\n{\n            'v': 9.35277271270752,\n            'f': \"9.35277271270752\",\n        },\nfalse,\n\"lot\",\n\"[MASK] of tourists loves to go for this tremendous fishing before their bbq meal , while the ukulhas island wills to provide fishing gears and all the required facilities on request .\",\n\"lot\",\n\"lots\",\n{\n            'v': 9.965821266174316,\n            'f': \"9.965821266174316\",\n        },\n{\n            'v': 10.507505416870115,\n            'f': \"10.507505416870115\",\n        },\nfalse],\n [{\n            'v': 11283,\n            'f': \"11283\",\n        },\n{\n            'v': 2,\n            'f': \"2\",\n        },\n\"major disturbances like a startup bubble burst leave behind an investment dry environment .\",\n\"major disturbances like a startup bubble burst [MASK] behind an investment dry environment .\",\n\"leave\",\n\"leaves\",\n{\n            'v': 4.611059665679932,\n            'f': \"4.611059665679932\",\n        },\n{\n            'v': 5.722336769104004,\n            'f': \"5.722336769104004\",\n        },\nfalse,\n\"disturbances\",\n\"major [MASK] like a startup bubble burst leave behind an investment dry environment .\",\n\"disturbances\",\n\"disturbance\",\n{\n            'v': -0.460705041885376,\n            'f': \"-0.460705041885376\",\n        },\n{\n            'v': 1.0977245569229126,\n            'f': \"1.0977245569229126\",\n        },\nfalse],\n [{\n            'v': 11746,\n            'f': \"11746\",\n        },\n{\n            'v': 2,\n            'f': \"2\",\n        },\n\"marine traffic has previously been observed to elicit responses in cetacean behaviours , but the cause and effects of these interactions has yet to be fully understood .\",\n\"marine traffic has previously been observed to elicit responses in cetacean behaviours , but the cause and effects of these interactions [MASK] yet to be fully understood .\",\n\"has\",\n\"have\",\n{\n            'v': 15.793819427490234,\n            'f': \"15.793819427490234\",\n        },\n{\n            'v': 18.28557014465332,\n            'f': \"18.28557014465332\",\n        },\nfalse,\n\"cause\",\n\"marine traffic has previously been observed to elicit responses in cetacean behaviours , but the [MASK] and effects of these interactions has yet to be fully understood .\",\n\"cause\",\n\"causes\",\n{\n            'v': 11.24096393585205,\n            'f': \"11.24096393585205\",\n        },\n{\n            'v': 14.017227172851562,\n            'f': \"14.017227172851562\",\n        },\nfalse],\n [{\n            'v': 12048,\n            'f': \"12048\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"merge with spirit of manila , the airlines has less than 10 destinations .\",\n\"merge with spirit of manila , the airlines [MASK] less than 10 destinations .\",\n\"has\",\n\"have\",\n{\n            'v': 7.397946834564209,\n            'f': \"7.397946834564209\",\n        },\n{\n            'v': 10.579889297485352,\n            'f': \"10.579889297485352\",\n        },\nfalse,\n\"airlines\",\n\"merge with spirit of manila , the [MASK] has less than 10 destinations .\",\n\"airlines\",\n\"airline\",\n{\n            'v': -3.281010150909424,\n            'f': \"-3.281010150909424\",\n        },\n{\n            'v': 1.118960738182068,\n            'f': \"1.118960738182068\",\n        },\nfalse],\n [{\n            'v': 12504,\n            'f': \"12504\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"most times these project feature a list of articles that need improvements .\",\n\"most times these project [MASK] a list of articles that need improvements .\",\n\"feature\",\n\"features\",\n{\n            'v': 7.738528251647949,\n            'f': \"7.738528251647949\",\n        },\n{\n            'v': 9.124256134033203,\n            'f': \"9.124256134033203\",\n        },\nfalse,\n\"project\",\n\"most times these [MASK] feature a list of articles that need improvements .\",\n\"project\",\n\"projects\",\n{\n            'v': 2.070530414581299,\n            'f': \"2.070530414581299\",\n        },\n{\n            'v': 7.927293300628662,\n            'f': \"7.927293300628662\",\n        },\nfalse],\n [{\n            'v': 12640,\n            'f': \"12640\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"my searches both here and across google have sort of turned up like a black hole pulled him off the face of the earth .\",\n\"my searches both here and across google [MASK] sort of turned up like a black hole pulled him off the face of the earth .\",\n\"have\",\n\"has\",\n{\n            'v': 5.239735126495361,\n            'f': \"5.239735126495361\",\n        },\n{\n            'v': 5.443183422088623,\n            'f': \"5.443183422088623\",\n        },\nfalse,\n\"searches\",\n\"my [MASK] both here and across google have sort of turned up like a black hole pulled him off the face of the earth .\",\n\"searches\",\n\"search\",\n{\n            'v': 4.134875774383545,\n            'f': \"4.134875774383545\",\n        },\n{\n            'v': 4.459184646606445,\n            'f': \"4.459184646606445\",\n        },\nfalse],\n [{\n            'v': 12671,\n            'f': \"12671\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"national central cities sphere of influence have great impact around the surrounding cities on modernizing and integrating services in fields such as infrastructure , finance , public education , social welfare , sanitation , business licensing and urban planning .\",\n\"national central cities sphere of influence [MASK] great impact around the surrounding cities on modernizing and integrating services in fields such as infrastructure , finance , public education , social welfare , sanitation , business licensing and urban planning .\",\n\"have\",\n\"has\",\n{\n            'v': 10.71315860748291,\n            'f': \"10.71315860748291\",\n        },\n{\n            'v': 14.515218734741213,\n            'f': \"14.515218734741213\",\n        },\nfalse,\n\"sphere\",\n\"national central cities [MASK] of influence have great impact around the surrounding cities on modernizing and integrating services in fields such as infrastructure , finance , public education , social welfare , sanitation , business licensing and urban planning .\",\n\"sphere\",\n\"spheres\",\n{\n            'v': 11.50922966003418,\n            'f': \"11.50922966003418\",\n        },\n{\n            'v': 13.59064769744873,\n            'f': \"13.59064769744873\",\n        },\nfalse],\n [{\n            'v': 12683,\n            'f': \"12683\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"native american disease and epidemics often leads individuals to think of alcoholism on tribal lands and surrounding areas .\",\n\"native american disease and epidemics often [MASK] individuals to think of alcoholism on tribal lands and surrounding areas .\",\n\"leads\",\n\"lead\",\n{\n            'v': 10.52984619140625,\n            'f': \"10.52984619140625\",\n        },\n{\n            'v': 13.794330596923828,\n            'f': \"13.794330596923828\",\n        },\nfalse,\n\"disease\",\n\"native american [MASK] and epidemics often leads individuals to think of alcoholism on tribal lands and surrounding areas .\",\n\"disease\",\n\"diseases\",\n{\n            'v': 11.4121732711792,\n            'f': \"11.4121732711792\",\n        },\n{\n            'v': 12.296198844909668,\n            'f': \"12.296198844909668\",\n        },\nfalse],\n [{\n            'v': 12741,\n            'f': \"12741\",\n        },\n{\n            'v': 4,\n            'f': \"4\",\n        },\n\"neither list of muslim historians , list of modern-day muslim scholars of islam , list of muslim philosophers , nor islamic studies seems a particularly good fit for the concept .\",\n\"neither list of muslim historians , list of modern-day muslim scholars of islam , list of muslim philosophers , nor islamic studies [MASK] a particularly good fit for the concept .\",\n\"seems\",\n\"seem\",\n{\n            'v': 8.503365516662598,\n            'f': \"8.503365516662598\",\n        },\n{\n            'v': 10.09550952911377,\n            'f': \"10.09550952911377\",\n        },\nfalse,\n\"list\",\n\"neither [MASK] of muslim historians , [MASK] of modern-day muslim scholars of islam , [MASK] of muslim philosophers , nor islamic studies seems a particularly good fit for the concept .\",\n\"list\",\n\"lists\",\n{\n            'v': 6.19161319732666,\n            'f': \"6.19161319732666\",\n        },\n{\n            'v': 7.250743389129639,\n            'f': \"7.250743389129639\",\n        },\nfalse],\n [{\n            'v': 12831,\n            'f': \"12831\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"new system posting of the trains has arrival .\",\n\"new system posting of the trains [MASK] arrival .\",\n\"has\",\n\"have\",\n{\n            'v': 0.3174472153186798,\n            'f': \"0.3174472153186798\",\n        },\n{\n            'v': 1.562584638595581,\n            'f': \"1.562584638595581\",\n        },\nfalse,\n\"posting\",\n\"new system [MASK] of the trains has arrival .\",\n\"posting\",\n\"postings\",\n{\n            'v': -0.3335655629634857,\n            'f': \"-0.3335655629634857\",\n        },\n{\n            'v': 1.8121790885925293,\n            'f': \"1.8121790885925293\",\n        },\nfalse],\n [{\n            'v': 12868,\n            'f': \"12868\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"no benefits to having this be a template instead of a category , a lot of redlinks , only sets up precedent clutter people like kant with a horde of boxes .\",\n\"no benefits to having this be a template instead of a category , a lot of redlinks , only [MASK] up precedent clutter people like kant with a horde of boxes .\",\n\"sets\",\n\"set\",\n{\n            'v': 5.310222148895264,\n            'f': \"5.310222148895264\",\n        },\n{\n            'v': 8.3071928024292,\n            'f': \"8.3071928024292\",\n        },\nfalse,\n\"benefits\",\n\"no [MASK] to having this be a template instead of a category , a lot of redlinks , only sets up precedent clutter people like kant with a horde of boxes .\",\n\"benefits\",\n\"benefit\",\n{\n            'v': 3.477786064147949,\n            'f': \"3.477786064147949\",\n        },\n{\n            'v': 7.850944519042969,\n            'f': \"7.850944519042969\",\n        },\nfalse],\n [{\n            'v': 13425,\n            'f': \"13425\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"offensive content discussion of the mechanics of tagging images ; the discussion of specific systems seems to have run into trouble with systems that require subjective judgements .\",\n\"offensive content discussion of the mechanics of tagging images ; the discussion of specific systems [MASK] to have run into trouble with systems that require subjective judgements .\",\n\"seems\",\n\"seem\",\n{\n            'v': 9.646284103393556,\n            'f': \"9.646284103393556\",\n        },\n{\n            'v': 9.87848663330078,\n            'f': \"9.87848663330078\",\n        },\nfalse,\n\"discussion\",\n\"offensive content [MASK] of the mechanics of tagging images ; the [MASK] of specific systems seems to have run into trouble with systems that require subjective judgements .\",\n\"discussion\",\n\"discussions\",\n{\n            'v': 4.329834938049316,\n            'f': \"4.329834938049316\",\n        },\n{\n            'v': 4.337092399597168,\n            'f': \"4.337092399597168\",\n        },\nfalse],\n [{\n            'v': 13434,\n            'f': \"13434\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"official portrait and for promotional purposes does not equal public domain .\",\n\"official portrait and for promotional purposes [MASK] not equal public domain .\",\n\"does\",\n\"do\",\n{\n            'v': 9.772114753723145,\n            'f': \"9.772114753723145\",\n        },\n{\n            'v': 10.908541679382324,\n            'f': \"10.908541679382324\",\n        },\nfalse,\n\"portrait\",\n\"official [MASK] and for promotional purposes does not equal public domain .\",\n\"portrait\",\n\"portraits\",\n{\n            'v': 3.9180846214294434,\n            'f': \"3.9180846214294434\",\n        },\n{\n            'v': 4.96876859664917,\n            'f': \"4.96876859664917\",\n        },\nfalse],\n [{\n            'v': 13827,\n            'f': \"13827\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"orientalizing period of vases ends in ancient greece .\",\n\"orientalizing period of vases [MASK] in ancient greece .\",\n\"ends\",\n\"end\",\n{\n            'v': 1.610647439956665,\n            'f': \"1.610647439956665\",\n        },\n{\n            'v': 1.868502140045166,\n            'f': \"1.868502140045166\",\n        },\nfalse,\n\"period\",\n\"orientalizing [MASK] of vases ends in ancient greece .\",\n\"period\",\n\"periods\",\n{\n            'v': 2.610722064971924,\n            'f': \"2.610722064971924\",\n        },\n{\n            'v': 3.09468412399292,\n            'f': \"3.09468412399292\",\n        },\nfalse],\n [{\n            'v': 13829,\n            'f': \"13829\",\n        },\n{\n            'v': 2,\n            'f': \"2\",\n        },\n\"original broadcasts and the sneak peek of the episode have the song in the end credits .\",\n\"original broadcasts and the sneak peek of the episode [MASK] the song in the end credits .\",\n\"have\",\n\"has\",\n{\n            'v': 7.304343700408935,\n            'f': \"7.304343700408935\",\n        },\n{\n            'v': 8.518545150756836,\n            'f': \"8.518545150756836\",\n        },\nfalse,\n\"broadcasts\",\n\"original [MASK] and the sneak peek of the episode have the song in the end credits .\",\n\"broadcasts\",\n\"broadcast\",\n{\n            'v': 5.449577331542969,\n            'f': \"5.449577331542969\",\n        },\n{\n            'v': 9.082113265991213,\n            'f': \"9.082113265991213\",\n        },\nfalse],\n [{\n            'v': 16297,\n            'f': \"16297\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"please note , that both rock bands ( plural form ) and bar band redirect to rock band as well .\",\n\"please note , that both rock bands ( plural form ) and bar [MASK] redirect to rock band as well .\",\n\"band\",\n\"bands\",\n{\n            'v': 8.959895133972168,\n            'f': \"8.959895133972168\",\n        },\n{\n            'v': 12.780133247375488,\n            'f': \"12.780133247375488\",\n        },\nfalse,\n\"bands\",\n\"please note , that both rock [MASK] ( plural form ) and bar band redirect to rock band as well .\",\n\"bands\",\n\"band\",\n{\n            'v': 7.991128921508789,\n            'f': \"7.991128921508789\",\n        },\n{\n            'v': 14.223956108093262,\n            'f': \"14.223956108093262\",\n        },\nfalse],\n [{\n            'v': 16348,\n            'f': \"16348\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"policy and guidelines follows consensus , often from afd .\",\n\"policy and guidelines [MASK] consensus , often from afd .\",\n\"follows\",\n\"follow\",\n{\n            'v': 3.2028393745422363,\n            'f': \"3.2028393745422363\",\n        },\n{\n            'v': 6.415107250213623,\n            'f': \"6.415107250213623\",\n        },\nfalse,\n\"policy\",\n\"[MASK] and guidelines follows consensus , often from afd .\",\n\"policy\",\n\"policies\",\n{\n            'v': 7.24785041809082,\n            'f': \"7.24785041809082\",\n        },\n{\n            'v': 7.332479000091553,\n            'f': \"7.332479000091553\",\n        },\nfalse],\n [{\n            'v': 17402,\n            'f': \"17402\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"revenge issue the meanest batch of essays cnf has ever published includes a post-divorce bonfire ; post-traumatic stress ; an assassination attempt ; a kidnapping plot ; dick cheney , and more .\",\n\"revenge issue the meanest batch of essays cnf has ever published [MASK] a post-divorce bonfire ; post-traumatic stress ; an assassination attempt ; a kidnapping plot ; dick cheney , and more .\",\n\"includes\",\n\"include\",\n{\n            'v': 6.756029605865479,\n            'f': \"6.756029605865479\",\n        },\n{\n            'v': 8.079351425170898,\n            'f': \"8.079351425170898\",\n        },\nfalse,\n\"issue\",\n\"revenge [MASK] the meanest batch of essays cnf has ever published includes a post-divorce bonfire ; post-traumatic stress ; an assassination attempt ; a kidnapping plot ; dick cheney , and more .\",\n\"issue\",\n\"issues\",\n{\n            'v': 1.8022288084030151,\n            'f': \"1.8022288084030151\",\n        },\n{\n            'v': 2.9064199924468994,\n            'f': \"2.9064199924468994\",\n        },\nfalse],\n [{\n            'v': 17513,\n            'f': \"17513\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"rugby league governing bodies in the united kingdom has four constituents .\",\n\"rugby league governing bodies in the united kingdom [MASK] four constituents .\",\n\"has\",\n\"have\",\n{\n            'v': 12.981082916259766,\n            'f': \"12.981082916259766\",\n        },\n{\n            'v': 14.803519248962402,\n            'f': \"14.803519248962402\",\n        },\nfalse,\n\"bodies\",\n\"rugby league governing [MASK] in the united kingdom has four constituents .\",\n\"bodies\",\n\"body\",\n{\n            'v': 13.927062034606934,\n            'f': \"13.927062034606934\",\n        },\n{\n            'v': 17.162811279296875,\n            'f': \"17.162811279296875\",\n        },\nfalse],\n [{\n            'v': 17685,\n            'f': \"17685\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"search for olmish yield 40 results , and a search for olmish people returns 6 results ( zero relevant ones outside wikipedia ) .\",\n\"search for olmish yield 40 results , and a search for olmish people [MASK] 6 results ( zero relevant ones outside wikipedia ) .\",\n\"returns\",\n\"return\",\n{\n            'v': 6.490913391113281,\n            'f': \"6.490913391113281\",\n        },\n{\n            'v': 7.074180603027344,\n            'f': \"7.074180603027344\",\n        },\nfalse,\n\"search\",\n\"[MASK] for olmish yield 40 results , and a [MASK] for olmish people returns 6 results ( zero relevant ones outside wikipedia ) .\",\n\"search\",\n\"searches\",\n{\n            'v': 4.2579498291015625,\n            'f': \"4.2579498291015625\",\n        },\n{\n            'v': 9.11375331878662,\n            'f': \"9.11375331878662\",\n        },\nfalse],\n [{\n            'v': 18831,\n            'f': \"18831\",\n        },\n{\n            'v': 3,\n            'f': \"3\",\n        },\n\"sounds and audio control panel applet represent the mme api of the sound card driver .\",\n\"sounds and audio control panel applet [MASK] the mme api of the sound card driver .\",\n\"represent\",\n\"represents\",\n{\n            'v': 5.198363780975342,\n            'f': \"5.198363780975342\",\n        },\n{\n            'v': 6.843611717224121,\n            'f': \"6.843611717224121\",\n        },\nfalse,\n\"sounds\",\n\"[MASK] and audio control panel applet represent the mme api of the sound card driver .\",\n\"sounds\",\n\"sound\",\n{\n            'v': 4.247204780578613,\n            'f': \"4.247204780578613\",\n        },\n{\n            'v': 6.569221496582031,\n            'f': \"6.569221496582031\",\n        },\nfalse],\n [{\n            'v': 18840,\n            'f': \"18840\",\n        },\n{\n            'v': 2,\n            'f': \"2\",\n        },\n\"source claims `` wilkes-barre high school '' , but of four high schools in that city , none have that name took art classes at the art students league of new york .\",\n\"source claims `` wilkes-barre high school '' , but of four high schools in that city , none [MASK] that name took art classes at the art students league of new york .\",\n\"have\",\n\"has\",\n{\n            'v': 6.131553649902344,\n            'f': \"6.131553649902344\",\n        },\n{\n            'v': 6.67499589920044,\n            'f': \"6.67499589920044\",\n        },\nfalse,\n\"source\",\n\"[MASK] claims `` wilkes-barre high school '' , but of four high schools in that city , none have that name took art classes at the art students league of new york .\",\n\"source\",\n\"sources\",\n{\n            'v': -1.8178681135177608,\n            'f': \"-1.8178681135177608\",\n        },\n{\n            'v': -1.372745394706726,\n            'f': \"-1.372745394706726\",\n        },\nfalse],\n [{\n            'v': 18921,\n            'f': \"18921\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"specialty finishes and its specialty marine decking unit provide industrial coating and marine decking services .\",\n\"specialty finishes and its specialty marine decking unit [MASK] industrial coating and marine decking services .\",\n\"provide\",\n\"provides\",\n{\n            'v': 12.78610610961914,\n            'f': \"12.78610610961914\",\n        },\n{\n            'v': 13.17575454711914,\n            'f': \"13.17575454711914\",\n        },\nfalse,\n\"finishes\",\n\"specialty [MASK] and its specialty marine decking unit provide industrial coating and marine decking services .\",\n\"finishes\",\n\"finish\",\n{\n            'v': 1.2392182350158691,\n            'f': \"1.2392182350158691\",\n        },\n{\n            'v': 1.2844069004058838,\n            'f': \"1.2844069004058838\",\n        },\nfalse],\n [{\n            'v': 19102,\n            'f': \"19102\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"stewie panics about the existence of the tooth fairy , brian and then chris fail to calm stewie down , then quagmire makes an attempt .\",\n\"stewie panics about the existence of the tooth fairy , brian and then chris [MASK] to calm stewie down , then quagmire makes an attempt .\",\n\"fail\",\n\"fails\",\n{\n            'v': 3.5772883892059326,\n            'f': \"3.5772883892059326\",\n        },\n{\n            'v': 4.269681930541992,\n            'f': \"4.269681930541992\",\n        },\nfalse,\n\"panics\",\n\"stewie [MASK] about the existence of the tooth fairy , brian and then chris fail to calm stewie down , then quagmire makes an attempt .\",\n\"panics\",\n\"panic\",\n{\n            'v': 0.4155803620815277,\n            'f': \"0.4155803620815277\",\n        },\n{\n            'v': 4.74290132522583,\n            'f': \"4.74290132522583\",\n        },\nfalse],\n [{\n            'v': 19419,\n            'f': \"19419\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"super version since their regular super turbo versions uses a different default color scheme .\",\n\"super version since their regular super turbo versions [MASK] a different default color scheme .\",\n\"uses\",\n\"use\",\n{\n            'v': 9.092540740966797,\n            'f': \"9.092540740966797\",\n        },\n{\n            'v': 9.794928550720217,\n            'f': \"9.794928550720217\",\n        },\nfalse,\n\"versions\",\n\"super version since their regular super turbo [MASK] uses a different default color scheme .\",\n\"versions\",\n\"version\",\n{\n            'v': 8.520002365112305,\n            'f': \"8.520002365112305\",\n        },\n{\n            'v': 9.920978546142578,\n            'f': \"9.920978546142578\",\n        },\nfalse],\n [{\n            'v': 19472,\n            'f': \"19472\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"support topic ban , with hopes that delta understands that i bear him absolutely no ill will because of our recent clash .\",\n\"support topic ban , with hopes that delta [MASK] that i bear him absolutely no ill will because of our recent clash .\",\n\"understands\",\n\"understand\",\n{\n            'v': 5.842806816101074,\n            'f': \"5.842806816101074\",\n        },\n{\n            'v': 6.4067888259887695,\n            'f': \"6.4067888259887695\",\n        },\nfalse,\n\"delta\",\n\"support topic ban , with hopes that [MASK] understands that i bear him absolutely no ill will because of our recent clash .\",\n\"delta\",\n\"deltum\",\n{\n            'v': -2.458537101745605,\n            'f': \"-2.458537101745605\",\n        },\n{\n            'v': 2.682741641998291,\n            'f': \"2.682741641998291\",\n        },\nfalse],\n [{\n            'v': 19690,\n            'f': \"19690\",\n        },\n{\n            'v': 2,\n            'f': \"2\",\n        },\n\"thanks for your answer , however the question remain .\",\n\"thanks for your answer , however the question [MASK] .\",\n\"remain\",\n\"remains\",\n{\n            'v': 8.028837203979492,\n            'f': \"8.028837203979492\",\n        },\n{\n            'v': 12.42752456665039,\n            'f': \"12.42752456665039\",\n        },\nfalse,\n\"question\",\n\"thanks for your answer , however the [MASK] remain .\",\n\"question\",\n\"questions\",\n{\n            'v': 7.197246074676514,\n            'f': \"7.197246074676514\",\n        },\n{\n            'v': 11.595948219299316,\n            'f': \"11.595948219299316\",\n        },\nfalse],\n [{\n            'v': 19992,\n            'f': \"19992\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the anglo-indian clubs of the time , most significantly , the calcutta naval volunteers club , now the calcutta rangers club deserve a big share of the credit .\",\n\"the anglo-indian clubs of the time , most significantly , the calcutta naval volunteers club , now the calcutta rangers club [MASK] a big share of the credit .\",\n\"deserve\",\n\"deserves\",\n{\n            'v': 2.6175553798675537,\n            'f': \"2.6175553798675537\",\n        },\n{\n            'v': 3.276566982269287,\n            'f': \"3.276566982269287\",\n        },\nfalse,\n\"clubs\",\n\"the anglo-indian [MASK] of the time , most significantly , the calcutta naval volunteers club , now the calcutta rangers club deserve a big share of the credit .\",\n\"clubs\",\n\"club\",\n{\n            'v': 7.674307346343994,\n            'f': \"7.674307346343994\",\n        },\n{\n            'v': 7.946598529815674,\n            'f': \"7.946598529815674\",\n        },\nfalse],\n [{\n            'v': 20089,\n            'f': \"20089\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the article mcleod 's daughters exists , and this information is already in that article to a great extent .\",\n\"the article mcleod 's daughters [MASK] , and this information is already in that article to a great extent .\",\n\"exists\",\n\"exist\",\n{\n            'v': 4.224635601043701,\n            'f': \"4.224635601043701\",\n        },\n{\n            'v': 5.766460418701172,\n            'f': \"5.766460418701172\",\n        },\nfalse,\n\"daughters\",\n\"the article mcleod 's [MASK] exists , and this information is already in that article to a great extent .\",\n\"daughters\",\n\"daughter\",\n{\n            'v': -1.3746588230133057,\n            'f': \"-1.3746588230133057\",\n        },\n{\n            'v': 3.183591842651367,\n            'f': \"3.183591842651367\",\n        },\nfalse],\n [{\n            'v': 21177,\n            'f': \"21177\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the crystal structures suggests that this family of proteins plays an important role in isoprenoid quinone metabolism or in transport or storage .\",\n\"the crystal structures suggests that this family of proteins [MASK] an important role in isoprenoid quinone metabolism or in transport or storage .\",\n\"plays\",\n\"play\",\n{\n            'v': 16.788448333740234,\n            'f': \"16.788448333740234\",\n        },\n{\n            'v': 16.906801223754883,\n            'f': \"16.906801223754883\",\n        },\nfalse,\n\"structures\",\n\"the crystal [MASK] suggests that this family of proteins plays an important role in isoprenoid quinone metabolism or in transport or storage .\",\n\"structures\",\n\"structure\",\n{\n            'v': 10.343047142028809,\n            'f': \"10.343047142028809\",\n        },\n{\n            'v': 17.17306900024414,\n            'f': \"17.17306900024414\",\n        },\nfalse],\n [{\n            'v': 21242,\n            'f': \"21242\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the day the x years runs out you can remember and have a guilty mind , and start using your ill-gotten goods .\",\n\"the day the x years [MASK] out you can remember and have a guilty mind , and start using your ill-gotten goods .\",\n\"runs\",\n\"run\",\n{\n            'v': 8.091266632080078,\n            'f': \"8.091266632080078\",\n        },\n{\n            'v': 12.226033210754396,\n            'f': \"12.226033210754396\",\n        },\nfalse,\n\"years\",\n\"the day the x [MASK] runs out you can remember and have a guilty mind , and start using your ill-gotten goods .\",\n\"years\",\n\"year\",\n{\n            'v': 1.6868951320648191,\n            'f': \"1.6868951320648191\",\n        },\n{\n            'v': 3.029676675796509,\n            'f': \"3.029676675796509\",\n        },\nfalse],\n [{\n            'v': 21872,\n            'f': \"21872\",\n        },\n{\n            'v': 3,\n            'f': \"3\",\n        },\n\"the famine codes and malnutrition section of the talk page have the relevant details .\",\n\"the famine codes and malnutrition section of the talk page [MASK] the relevant details .\",\n\"have\",\n\"has\",\n{\n            'v': 8.120706558227539,\n            'f': \"8.120706558227539\",\n        },\n{\n            'v': 9.864959716796877,\n            'f': \"9.864959716796877\",\n        },\nfalse,\n\"codes\",\n\"the famine [MASK] and malnutrition section of the talk page have the relevant details .\",\n\"codes\",\n\"code\",\n{\n            'v': 2.051108360290528,\n            'f': \"2.051108360290528\",\n        },\n{\n            'v': 2.7632384300231934,\n            'f': \"2.7632384300231934\",\n        },\nfalse],\n [{\n            'v': 22005,\n            'f': \"22005\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the first and last tracks on andrew bird 's album noble beast form a palindrome ( `` oh no '' and `` on ho !\",\n\"the first and last tracks on andrew bird 's album noble beast [MASK] a palindrome ( `` oh no '' and `` on ho !\",\n\"form\",\n\"forms\",\n{\n            'v': 3.2411999702453613,\n            'f': \"3.2411999702453613\",\n        },\n{\n            'v': 3.752320051193237,\n            'f': \"3.752320051193237\",\n        },\nfalse,\n\"tracks\",\n\"the first and last [MASK] on andrew bird 's album noble beast form a palindrome ( `` oh no '' and `` on ho !\",\n\"tracks\",\n\"track\",\n{\n            'v': 6.065039157867432,\n            'f': \"6.065039157867432\",\n        },\n{\n            'v': 6.477729797363281,\n            'f': \"6.477729797363281\",\n        },\nfalse],\n [{\n            'v': 22117,\n            'f': \"22117\",\n        },\n{\n            'v': 3,\n            'f': \"3\",\n        },\n\"the fishery and aquaculture sectors in chile apart from salmon and trout chilean aquaculture also grow turbots and molluscs in particular mytilus chilensis , northern scallops , pacific oysters , aulacomya ater ( a giant mussel ) , red abalone , and chilean oysters .\",\n\"the fishery and aquaculture sectors in chile apart from salmon and trout chilean aquaculture also [MASK] turbots and molluscs in particular mytilus chilensis , northern scallops , pacific oysters , aulacomya ater ( a giant mussel ) , red abalone , and chilean oysters .\",\n\"grow\",\n\"grows\",\n{\n            'v': 4.979804515838623,\n            'f': \"4.979804515838623\",\n        },\n{\n            'v': 6.990700721740723,\n            'f': \"6.990700721740723\",\n        },\nfalse,\n\"sectors\",\n\"the fishery and aquaculture [MASK] in chile apart from salmon and trout chilean aquaculture also grow turbots and molluscs in particular mytilus chilensis , northern scallops , pacific oysters , aulacomya ater ( a giant mussel ) , red abalone , and chilean oysters .\",\n\"sectors\",\n\"sector\",\n{\n            'v': 8.701552391052246,\n            'f': \"8.701552391052246\",\n        },\n{\n            'v': 9.661343574523926,\n            'f': \"9.661343574523926\",\n        },\nfalse],\n [{\n            'v': 22155,\n            'f': \"22155\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the flower gardens hirschstetten , with an office in essling , kultivationsbetriebe serve primarily as the city office .\",\n\"the flower gardens hirschstetten , with an office in essling , kultivationsbetriebe [MASK] primarily as the city office .\",\n\"serve\",\n\"serves\",\n{\n            'v': 9.987278938293455,\n            'f': \"9.987278938293455\",\n        },\n{\n            'v': 12.017707824707031,\n            'f': \"12.017707824707031\",\n        },\nfalse,\n\"gardens\",\n\"the flower [MASK] hirschstetten , with an office in essling , kultivationsbetriebe serve primarily as the city office .\",\n\"gardens\",\n\"garden\",\n{\n            'v': 9.761482238769531,\n            'f': \"9.761482238769531\",\n        },\n{\n            'v': 11.645044326782228,\n            'f': \"11.645044326782228\",\n        },\nfalse],\n [{\n            'v': 22360,\n            'f': \"22360\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the genetic mutations leading to light skin , though different among east asians and western europeans , suggest the two groups experienced a similar selective pressure due to settlement in northern latitudes .\",\n\"the genetic mutations leading to light skin , though different among east asians and western europeans , [MASK] the two groups experienced a similar selective pressure due to settlement in northern latitudes .\",\n\"suggest\",\n\"suggests\",\n{\n            'v': 12.126016616821287,\n            'f': \"12.126016616821287\",\n        },\n{\n            'v': 12.61373805999756,\n            'f': \"12.61373805999756\",\n        },\nfalse,\n\"mutations\",\n\"the genetic [MASK] leading to light skin , though different among east asians and western europeans , suggest the two groups experienced a similar selective pressure due to settlement in northern latitudes .\",\n\"mutations\",\n\"mutation\",\n{\n            'v': 9.774313926696776,\n            'f': \"9.774313926696776\",\n        },\n{\n            'v': 9.832053184509276,\n            'f': \"9.832053184509276\",\n        },\nfalse],\n [{\n            'v': 22483,\n            'f': \"22483\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the heaps of mining waste remain , contaminated with lead , and on which little will grow .\",\n\"the heaps of mining waste [MASK] , contaminated with lead , and on which little will grow .\",\n\"remain\",\n\"remains\",\n{\n            'v': 5.011368274688721,\n            'f': \"5.011368274688721\",\n        },\n{\n            'v': 6.484438896179199,\n            'f': \"6.484438896179199\",\n        },\nfalse,\n\"heaps\",\n\"the [MASK] of mining waste remain , contaminated with lead , and on which little will grow .\",\n\"heaps\",\n\"heap\",\n{\n            'v': 1.8925843238830569,\n            'f': \"1.8925843238830569\",\n        },\n{\n            'v': 8.23051643371582,\n            'f': \"8.23051643371582\",\n        },\nfalse],\n [{\n            'v': 22700,\n            'f': \"22700\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the included application programming interfaces ( apis ) allows to configure and extend the functionalities of websites .\",\n\"the included application programming interfaces ( apis ) [MASK] to configure and extend the functionalities of websites .\",\n\"allows\",\n\"allow\",\n{\n            'v': 7.818078517913818,\n            'f': \"7.818078517913818\",\n        },\n{\n            'v': 9.803810119628906,\n            'f': \"9.803810119628906\",\n        },\nfalse,\n\"interfaces\",\n\"the included application programming [MASK] ( apis ) allows to configure and extend the functionalities of websites .\",\n\"interfaces\",\n\"interface\",\n{\n            'v': 7.074940204620361,\n            'f': \"7.074940204620361\",\n        },\n{\n            'v': 11.613018035888672,\n            'f': \"11.613018035888672\",\n        },\nfalse],\n [{\n            'v': 22939,\n            'f': \"22939\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the jets , as a member of the nfl have access to huge quantities of statistics and league historical information .\",\n\"the jets , as a member of the nfl [MASK] access to huge quantities of statistics and league historical information .\",\n\"have\",\n\"has\",\n{\n            'v': 11.826630592346191,\n            'f': \"11.826630592346191\",\n        },\n{\n            'v': 13.351736068725586,\n            'f': \"13.351736068725586\",\n        },\nfalse,\n\"jets\",\n\"the [MASK] , as a member of the nfl have access to huge quantities of statistics and league historical information .\",\n\"jets\",\n\"jet\",\n{\n            'v': -2.128173589706421,\n            'f': \"-2.128173589706421\",\n        },\n{\n            'v': -1.8322508335113523,\n            'f': \"-1.8322508335113523\",\n        },\nfalse],\n [{\n            'v': 22969,\n            'f': \"22969\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the kinds of behavior ttt is showing have no place on a collaborative project .\",\n\"the kinds of behavior ttt is showing [MASK] no place on a collaborative project .\",\n\"have\",\n\"has\",\n{\n            'v': 8.65194034576416,\n            'f': \"8.65194034576416\",\n        },\n{\n            'v': 9.137454986572266,\n            'f': \"9.137454986572266\",\n        },\nfalse,\n\"kinds\",\n\"the [MASK] of behavior ttt is showing have no place on a collaborative project .\",\n\"kinds\",\n\"kind\",\n{\n            'v': 7.217334747314453,\n            'f': \"7.217334747314453\",\n        },\n{\n            'v': 9.870264053344728,\n            'f': \"9.870264053344728\",\n        },\nfalse],\n [{\n            'v': 24103,\n            'f': \"24103\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the notability guideline for films specifically excludes the internet movie database and similar sites from the definition of non-trivial coverage , because they indiscriminately list almost every film .\",\n\"the notability guideline for films specifically [MASK] the internet movie database and similar sites from the definition of non-trivial coverage , because they indiscriminately list almost every film .\",\n\"excludes\",\n\"exclude\",\n{\n            'v': 3.449037313461304,\n            'f': \"3.449037313461304\",\n        },\n{\n            'v': 10.05422592163086,\n            'f': \"10.05422592163086\",\n        },\nfalse,\n\"guideline\",\n\"the notability [MASK] for films specifically excludes the internet movie database and similar sites from the definition of non-trivial coverage , because they indiscriminately list almost every film .\",\n\"guideline\",\n\"guidelines\",\n{\n            'v': 3.5881965160369877,\n            'f': \"3.5881965160369877\",\n        },\n{\n            'v': 5.751475811004639,\n            'f': \"5.751475811004639\",\n        },\nfalse],\n [{\n            'v': 24109,\n            'f': \"24109\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the notched edges and the mauve colour recall the symbol flower of hong kong , the bauhinia blakeana , while the colour red is seen as particularly auspicious .\",\n\"the notched edges and the mauve colour [MASK] the symbol flower of hong kong , the bauhinia blakeana , while the colour red is seen as particularly auspicious .\",\n\"recall\",\n\"recalls\",\n{\n            'v': 7.156277179718018,\n            'f': \"7.156277179718018\",\n        },\n{\n            'v': 7.224477291107178,\n            'f': \"7.224477291107178\",\n        },\nfalse,\n\"edges\",\n\"the notched [MASK] and the mauve colour recall the symbol flower of hong kong , the bauhinia blakeana , while the colour red is seen as particularly auspicious .\",\n\"edges\",\n\"edge\",\n{\n            'v': 8.682877540588379,\n            'f': \"8.682877540588379\",\n        },\n{\n            'v': 9.610868453979492,\n            'f': \"9.610868453979492\",\n        },\nfalse],\n [{\n            'v': 24397,\n            'f': \"24397\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the original and main title card used for later seasons has a portion of both towers visible , yet they are not very obvious .\",\n\"the original and main title card used for later seasons [MASK] a portion of both towers visible , yet they are not very obvious .\",\n\"has\",\n\"have\",\n{\n            'v': 11.991801261901855,\n            'f': \"11.991801261901855\",\n        },\n{\n            'v': 12.500773429870604,\n            'f': \"12.500773429870604\",\n        },\nfalse,\n\"card\",\n\"the original and main title [MASK] used for later seasons has a portion of both towers visible , yet they are not very obvious .\",\n\"card\",\n\"cards\",\n{\n            'v': 7.291306972503662,\n            'f': \"7.291306972503662\",\n        },\n{\n            'v': 8.498577117919922,\n            'f': \"8.498577117919922\",\n        },\nfalse],\n [{\n            'v': 24793,\n            'f': \"24793\",\n        },\n{\n            'v': 3,\n            'f': \"3\",\n        },\n\"the praises of guru sung in a devoted and emotional manner in this book make baba jiwan singh a poet of the rank of bhatt poets in guru granth sahib .\",\n\"the praises of guru sung in a devoted and emotional manner in this book [MASK] baba jiwan singh a poet of the rank of bhatt poets in guru granth sahib .\",\n\"make\",\n\"makes\",\n{\n            'v': 5.824635982513428,\n            'f': \"5.824635982513428\",\n        },\n{\n            'v': 6.729640007019043,\n            'f': \"6.729640007019043\",\n        },\nfalse,\n\"praises\",\n\"the [MASK] of guru sung in a devoted and emotional manner in this book make baba jiwan singh a poet of the rank of bhatt poets in guru granth sahib .\",\n\"praises\",\n\"praise\",\n{\n            'v': 3.1302897930145264,\n            'f': \"3.1302897930145264\",\n        },\n{\n            'v': 10.659515380859377,\n            'f': \"10.659515380859377\",\n        },\nfalse],\n [{\n            'v': 24977,\n            'f': \"24977\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"the qrs complexes with a regular r to r interval represent the second rhythm .\",\n\"the qrs complexes with a regular r to r interval [MASK] the second rhythm .\",\n\"represent\",\n\"represents\",\n{\n            'v': 6.195212364196777,\n            'f': \"6.195212364196777\",\n        },\n{\n            'v': 6.414074420928955,\n            'f': \"6.414074420928955\",\n        },\nfalse,\n\"complexes\",\n\"the qrs [MASK] with a regular r to r interval represent the second rhythm .\",\n\"complexes\",\n\"complex\",\n{\n            'v': 1.384735941886902,\n            'f': \"1.384735941886902\",\n        },\n{\n            'v': 3.3334145545959477,\n            'f': \"3.3334145545959477\",\n        },\nfalse],\n [{\n            'v': 25796,\n            'f': \"25796\",\n        },\n{\n            'v': 3,\n            'f': \"3\",\n        },\n\"these shapes with five-fold symmetry repeat periodically , although the space between them contain other shapes that do not have a periodic pattern .\",\n\"these shapes with five-fold symmetry repeat periodically , although the space between them [MASK] other shapes that do not have a periodic pattern .\",\n\"contain\",\n\"contains\",\n{\n            'v': 7.736605167388916,\n            'f': \"7.736605167388916\",\n        },\n{\n            'v': 12.93440055847168,\n            'f': \"12.93440055847168\",\n        },\nfalse,\n\"space\",\n\"these shapes with five-fold symmetry repeat periodically , although the [MASK] between them contain other shapes that do not have a periodic pattern .\",\n\"space\",\n\"spaces\",\n{\n            'v': 10.10769271850586,\n            'f': \"10.10769271850586\",\n        },\n{\n            'v': 12.809431076049805,\n            'f': \"12.809431076049805\",\n        },\nfalse],\n [{\n            'v': 27292,\n            'f': \"27292\",\n        },\n{\n            'v': 2,\n            'f': \"2\",\n        },\n\"the writers of the pop song whiter shade of pale claim the inspiration came from bach 's air on a g string cigar advertisement .\",\n\"the writers of the pop song whiter shade of pale [MASK] the inspiration came from bach 's air on a g string cigar advertisement .\",\n\"claim\",\n\"claims\",\n{\n            'v': 2.449135065078736,\n            'f': \"2.449135065078736\",\n        },\n{\n            'v': 3.4616641998291016,\n            'f': \"3.4616641998291016\",\n        },\nfalse,\n\"writers\",\n\"the [MASK] of the pop song whiter shade of pale claim the inspiration came from bach 's air on a g string cigar advertisement .\",\n\"writers\",\n\"writer\",\n{\n            'v': 2.69872522354126,\n            'f': \"2.69872522354126\",\n        },\n{\n            'v': 4.575622081756592,\n            'f': \"4.575622081756592\",\n        },\nfalse],\n [{\n            'v': 27915,\n            'f': \"27915\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"this whole article is written with the basic , and false , assumption the groups seeks what it claims to seek .\",\n\"this whole article is written with the basic , and false , assumption the groups [MASK] what it claims to seek .\",\n\"seeks\",\n\"seek\",\n{\n            'v': 7.887539386749268,\n            'f': \"7.887539386749268\",\n        },\n{\n            'v': 9.920659065246582,\n            'f': \"9.920659065246582\",\n        },\nfalse,\n\"groups\",\n\"this whole article is written with the basic , and false , assumption the [MASK] seeks what it claims to seek .\",\n\"groups\",\n\"group\",\n{\n            'v': 1.0379319190979004,\n            'f': \"1.0379319190979004\",\n        },\n{\n            'v': 7.843909740447998,\n            'f': \"7.843909740447998\",\n        },\nfalse],\n [{\n            'v': 28050,\n            'f': \"28050\",\n        },\n{\n            'v': 2,\n            'f': \"2\",\n        },\n\"through its newsletter , direct email and social media campaigns , first peoples worldwide provides information to individuals on how they can help protect indigenous peoples and foster the development of indigenous communities .\",\n\"through its newsletter , direct email and social media campaigns , first peoples worldwide [MASK] information to individuals on how they can help protect indigenous peoples and foster the development of indigenous communities .\",\n\"provides\",\n\"provide\",\n{\n            'v': 14.028714179992676,\n            'f': \"14.028714179992676\",\n        },\n{\n            'v': 15.61001205444336,\n            'f': \"15.61001205444336\",\n        },\nfalse,\n\"peoples\",\n\"through its newsletter , direct email and social media campaigns , first [MASK] worldwide provides information to individuals on how they can help protect indigenous [MASK] and foster the development of indigenous communities .\",\n\"peoples\",\n\"people\",\n{\n            'v': 0.5845378637313843,\n            'f': \"0.5845378637313843\",\n        },\n{\n            'v': 2.6215384006500244,\n            'f': \"2.6215384006500244\",\n        },\nfalse],\n [{\n            'v': 28211,\n            'f': \"28211\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"today two set of bridges known as san antonio and dos hermanos connects the adjoining areas of old san juan and condado .\",\n\"today two set of bridges known as san antonio and dos hermanos [MASK] the adjoining areas of old san juan and condado .\",\n\"connects\",\n\"connect\",\n{\n            'v': 7.493304252624512,\n            'f': \"7.493304252624512\",\n        },\n{\n            'v': 10.425220489501951,\n            'f': \"10.425220489501951\",\n        },\nfalse,\n\"set\",\n\"today two [MASK] of bridges known as san antonio and dos hermanos connects the adjoining areas of old san juan and condado .\",\n\"set\",\n\"sets\",\n{\n            'v': 10.506369590759276,\n            'f': \"10.506369590759276\",\n        },\n{\n            'v': 13.608819007873535,\n            'f': \"13.608819007873535\",\n        },\nfalse],\n [{\n            'v': 28778,\n            'f': \"28778\",\n        },\n{\n            'v': 3,\n            'f': \"3\",\n        },\n\"update that fixes issues with limited games and blow your e-fuses ( makes it non-jtagable ) however this is the last update you could have and still be able to jtag .\",\n\"update that fixes issues with limited games and blow your e-fuses ( [MASK] it non-jtagable ) however this is the last update you could have and still be able to jtag .\",\n\"makes\",\n\"make\",\n{\n            'v': 13.07426643371582,\n            'f': \"13.07426643371582\",\n        },\n{\n            'v': 15.345673561096191,\n            'f': \"15.345673561096191\",\n        },\nfalse,\n\"fixes\",\n\"update that [MASK] issues with limited games and blow your e-fuses ( makes it non-jtagable ) however this is the last update you could have and still be able to jtag .\",\n\"fixes\",\n\"fix\",\n{\n            'v': 0.9528560638427734,\n            'f': \"0.9528560638427734\",\n        },\n{\n            'v': 7.454342365264893,\n            'f': \"7.454342365264893\",\n        },\nfalse],\n [{\n            'v': 28933,\n            'f': \"28933\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"usually these group of people goes around at night in large groups up to 100 bikes with loud mufflers causing noise pollution .\",\n\"usually these group of people [MASK] around at night in large groups up to 100 bikes with loud mufflers causing noise pollution .\",\n\"goes\",\n\"go\",\n{\n            'v': 6.57088041305542,\n            'f': \"6.57088041305542\",\n        },\n{\n            'v': 10.423775672912598,\n            'f': \"10.423775672912598\",\n        },\nfalse,\n\"group\",\n\"usually these [MASK] of people goes around at night in large [MASK]s up to 100 bikes with loud mufflers causing noise pollution .\",\n\"group\",\n\"groups\",\n{\n            'v': 12.92775535583496,\n            'f': \"12.92775535583496\",\n        },\n{\n            'v': 13.05739688873291,\n            'f': \"13.05739688873291\",\n        },\nfalse],\n [{\n            'v': 29236,\n            'f': \"29236\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"weak keep it is referred to a lot by various websites ( mainly pot alternative lifestyle related , as well as forums - gets 129 , 000 hits on google .\",\n\"weak keep it is referred to a lot by various websites ( mainly pot alternative lifestyle related , as well as forums - [MASK] 129 , 000 hits on google .\",\n\"gets\",\n\"get\",\n{\n            'v': 4.836196422576904,\n            'f': \"4.836196422576904\",\n        },\n{\n            'v': 6.246789455413818,\n            'f': \"6.246789455413818\",\n        },\nfalse,\n\"lifestyle\",\n\"weak keep it is referred to a lot by various websites ( mainly pot alternative [MASK] related , as well as forums - gets 129 , 000 hits on google .\",\n\"lifestyle\",\n\"lifestyles\",\n{\n            'v': 0.7587180137634277,\n            'f': \"0.7587180137634277\",\n        },\n{\n            'v': 0.9900938272476196,\n            'f': \"0.9900938272476196\",\n        },\nfalse],\n [{\n            'v': 29239,\n            'f': \"29239\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"weak oppose slanted posts , glare on the hat , some noise on the clothes and looks excessively noise removed .\",\n\"weak oppose slanted posts , glare on the hat , some noise on the clothes and [MASK] excessively noise removed .\",\n\"looks\",\n\"look\",\n{\n            'v': 3.129375457763672,\n            'f': \"3.129375457763672\",\n        },\n{\n            'v': 3.4307432174682617,\n            'f': \"3.4307432174682617\",\n        },\nfalse,\n\"weak\",\n\"[MASK] oppose slanted posts , glare on the hat , some noise on the clothes and looks excessively noise removed .\",\n\"weak\",\n\"weaks\",\n{\n            'v': -1.4800485372543335,\n            'f': \"-1.4800485372543335\",\n        },\n{\n            'v': 2.497913360595703,\n            'f': \"2.497913360595703\",\n        },\nfalse],\n [{\n            'v': 29507,\n            'f': \"29507\",\n        },\n{\n            'v': 1,\n            'f': \"1\",\n        },\n\"while in skinner 's office , the camera angles and position of the characters reveals that the scene is intended to be a recreation of leonardo da vinci 's famous painting the last supper .\",\n\"while in skinner 's office , the camera angles and position of the characters [MASK] that the scene is intended to be a recreation of leonardo da vinci 's famous painting the last supper .\",\n\"reveals\",\n\"reveal\",\n{\n            'v': 10.740784645080566,\n            'f': \"10.740784645080566\",\n        },\n{\n            'v': 12.171119689941406,\n            'f': \"12.171119689941406\",\n        },\nfalse,\n\"angles\",\n\"while in skinner 's office , the camera [MASK] and position of the characters reveals that the scene is intended to be a recreation of leonardo da vinci 's famous painting the last supper .\",\n\"angles\",\n\"angle\",\n{\n            'v': 11.892818450927734,\n            'f': \"11.892818450927734\",\n        },\n{\n            'v': 13.31429386138916,\n            'f': \"13.31429386138916\",\n        },\nfalse]],\n        columns: [[\"number\", \"Unnamed: 0\"], [\"number\", \"attractors\"], [\"string\", \"full_sentence\"], [\"string\", \"verb_masked_sentence\"], [\"string\", \"verb_true_label\"], [\"string\", \"verb_false_label\"], [\"number\", \"verb_true_label_p\"], [\"number\", \"verb_false_label_p\"], [\"string\", \"verb_is_accurate\"], [\"string\", \"subject\"], [\"string\", \"subject_masked_sentence\"], [\"string\", \"subject_true_label\"], [\"string\", \"subject_false_label\"], [\"number\", \"subject_true_label_p\"], [\"number\", \"subject_false_label_p\"], [\"string\", \"subject_is_accurate\"]],\n        columnOptions: [],\n        rowsPerPage: 5,\n        helpUrl: \"https://colab.research.google.com/notebooks/data_table.ipynb\",\n        suppressOutputScrolling: true,\n        minimumWidth: undefined,\n      });\n    ",
            "text/plain": [
              "<google.colab.data_table.DataTable object>"
            ]
          },
          "metadata": {},
          "execution_count": 15
        }
      ],
      "source": [
        "from google.colab import data_table\n",
        "data_table.enable_dataframe_formatter()\n",
        "\n",
        "print(\"Sentences where both verb and subject is incorrectly chosen by the model:\")\n",
        "data = filtered_results[(filtered_results['subject_is_accurate'] == False) & (filtered_results['verb_is_accurate'] == False)]\n",
        "data_table.DataTable(data, include_index=False, num_rows_per_page=5)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 776
        },
        "id": "FT0P4cwSLPHH",
        "outputId": "5c539b1d-e807-41ff-9356-6ecf50547fb6"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjAAAAFWCAYAAAB3tKe1AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABdc0lEQVR4nO3dd3xT5f4H8E+SZnWkg7bQFmhpy55SQVYplr0UVChwLzIcKCh6f1dwXWU4uYiCbFBRERUBxQuyEWSIikzZs2xoS+neyfP749hDQzpSaHuS5vN+vfqCnpwk35yenHzynOd5jkoIIUBERETkRNRKF0BERERUXgwwRERE5HQYYIiIiMjpMMAQERGR02GAISIiIqfDAENEREROhwGGiIiInA4DDBERETkdBhgiIiJyOtUqwHz++edQqVSIj4+v8udWqVR47rnnqvx5STlhYWEYOXKk0mUoztG2w9KlS9GoUSNotVr4+PgoXY5dJk+eDJVKhaSkpEp7jvj4eKhUKnz++eeV9hwlPecHH3xQ5rqF24BsOeK26dKlC5o1a6ZoDXcVYI4ePYp//vOfCAkJgV6vR3BwMP7xj3/g6NGjFV2fy7p69SomT56MgwcPKl2KS/v1118xefJkpKSkKF2K0zh27BgmT56syBeJEydOYOTIkYiIiMDixYuxaNGiKq+BHNO6deswefJkpcugClTuAPP999+jdevW2Lp1K0aNGoV58+bhiSeewLZt29C6dWv88MMPlVGny7l69SqmTJnCAKOwX3/9FVOmTCk2wJw8eRKLFy+u+qIc3LFjxzBlyhRFAsz27dthsVgwa9YsjBw5EoMHD67yGuju/Oc//0F2dnalPf66deswZcqUSnt8qnpu5Vn57NmzGD58OMLDw7Fjxw4EBATIt73wwguIjo7G8OHDcfjwYYSHh1d4sSXJzMyEh4dHlT0fEQDo9XqlSwAAWCwW5OXlwWAwKF2K4hISEgDAaU4d0W1ubm5wcyvXRxIpwJE+b8vVAjN9+nRkZWVh0aJFVuEFAPz9/bFw4UJkZmbiv//9LwBg5cqVUKlU+OWXX2wea+HChVCpVDhy5Ii87MSJE3jsscfg5+cHg8GA+++/H//73/+s7lfYz+WXX37B2LFjERgYiNq1a5dY848//oi+ffsiODgYer0eEREReOutt2A2m63WKzyft2/fPnTo0AFGoxH16tXDggULyrOJsGzZMjRs2BAGgwFRUVHYsWOHzTpXrlzB6NGjUbNmTej1ejRt2hSfffaZfPv27dvRpk0bAMCoUaOgUqnkc9cff/wxNBqNVYvAjBkzoFKp8H//93/yMrPZDC8vL7z88svyMovFgpkzZ6Jp06YwGAyoWbMmxowZg1u3btnUuH79ekRHR8PDwwNeXl7o27evzSnCkSNHwtPTE1euXMGAAQPg6emJgIAAvPTSSzbbtzhhYWHo168fdu3ahbZt28JgMCA8PBxffvmlzbopKSl48cUXUadOHej1ekRGRmLatGmwWCxW6928eRPDhw+HyWSCj48PRowYgUOHDtmc+z98+DBGjhyJ8PBwGAwG1KpVC6NHj8bNmzfldSZPnowJEyYAAOrVqyf/HQpbFor2/fjzzz+hUqnwxRdf2NS+ceNGqFQqrF27Vl5W1j5QmsL+VsuWLUPTpk2h1+uxYcOGcj3u7Nmz0bRpU7i7u8PX1xf3338/vv76a/n2kSNHIiwszOZ+ZZ2L//zzzzFo0CAAwIMPPihvs+3bt8vbqWfPnvD395ffY6NHj7brdc+bN09+vcHBwRg3bpzV+yAsLAyTJk0CAAQEBEClUpV6yqBw/7148SL69esHT09PhISEYO7cuQCAv/76C7GxsfDw8EBoaKjV9gGA5ORkvPTSS2jevDk8PT1hMpnQu3dvHDp0yOa5ytrexblw4QIiIyPRrFkz3LhxA4D974OUlBSMHDkS3t7e8vvA3tOg+fn5mDJlCurXrw+DwYAaNWqgU6dO2Lx5s7xOly5d0KVLF5v7lrTfAMBHH32E0NBQGI1GxMTEWB37gZL3ra+++gpRUVEwGo3w8/PDkCFDcOnSJZv1fv/9d/Tp0we+vr7w8PBAixYtMGvWLLmuwr9r4T5Z9Lm+/fZbREVFwcvLCyaTCc2bN5fvW5oPPvgAHTp0QI0aNWA0GhEVFYWVK1farFf4nl29ejWaNWsmvzcL37dF7dq1C23atIHBYEBERAQWLlxYZh0A8Nxzz8HT0xNZWVk2tw0dOhS1atWyOi6X5xh/9uxZ9OnTB15eXvjHP/5htY49n5n27P8nTpzAxYsX7XqtMlEOwcHBIiwsrNR1wsLCRO3atYUQQmRlZQlPT08xduxYm/UefPBB0bRpU/n3I0eOCG9vb9GkSRMxbdo0MWfOHNG5c2ehUqnE999/L6+3ZMkSAUA0adJExMTEiNmzZ4v333/f6rbz58/L6w8YMEAMHjxYTJ8+XcyfP18MGjRIABAvvfSSVT0xMTEiODhYBAYGiueee058/PHHolOnTgKA+PTTT8vcNgBEs2bNhL+/v5g6daqYNm2aCA0NFUajUfz111/yetevXxe1a9cWderUEVOnThXz588XDz30kAAgPvroI3mdqVOnCgDi6aefFkuXLhVLly4VZ8+eFfv37xcAxJo1a+THfPjhh4VarRb333+/vGzv3r0CgFi7dq287MknnxRubm7iqaeeEgsWLBAvv/yy8PDwEG3atBF5eXnyel9++aVQqVSiV69eYvbs2WLatGkiLCxM+Pj4WG3bESNGCIPBIJo2bSpGjx4t5s+fLx599FEBQMybN6/MbRYaGioaNmwoatasKV577TUxZ84c0bp1a6FSqcSRI0fk9TIzM0WLFi1EjRo1xGuvvSYWLFggHn/8caFSqcQLL7wgr2c2m0X79u2FRqMRzz33nJgzZ47o3r27aNmypQAglixZIq/7wQcfiOjoaDF16lSxaNEi8cILLwij0Sjatm0rLBaLEEKIQ4cOiaFDh8p/m8K/Q0ZGhlz/iBEj5McMDw8Xffr0sXmdo0aNEr6+vvI2tmcfKA0A0bhxYxEQECCmTJki5s6dKw4cOGD34y5atEgAEI899phYuHChmDVrlnjiiSfE+PHj5XVGjBghQkNDbZ570qRJ4s7DRtHtcPbsWTF+/HgBQLz22mvyNrt+/bq4ceOG8PX1FQ0aNBDTp08XixcvFq+//rpo3Lhxma+58Hm7desmZs+eLZ577jmh0Wis9t0ffvhBDBw4UAAQ8+fPF0uXLhWHDh0q8TEL998mTZqIZ555RsydO1d06NBB3leCg4PFhAkTxOzZs0XTpk2FRqMR586dk++/d+9eERERIV555RWxcOFCMXXqVBESEiK8vb3FlStXyrW9C19fYmKiEEKIM2fOiLp164pWrVrJy+x9H1gsFtG5c2ehVqvF2LFjxezZs0VsbKxo0aKFzfugOK+99ppQqVTiqaeeEosXLxYzZswQQ4cOlY+zQkjHy5iYmGK3adH95vz58wKAaN68uQgLCxPTpk0TU6ZMEX5+fiIgIEBcv37dZhsU9fbbbwuVSiXi4uLEvHnzxJQpU4S/v78ICwsTt27dktfbtGmT0Ol0IjQ0VEyaNEnMnz9fjB8/XnTr1k0IIcSvv/4qunfvLgDI++TSpUvl+wIQXbt2FXPnzhVz584Vzz33nBg0aFCp20kIIWrXri3Gjh0r5syZIz788EPRtm1bm+OuENJ7tmXLliIoKEi89dZbYubMmSI8PFy4u7uLpKQkeb3Dhw8Lo9Eo6tatK9577z3x1ltviZo1a8p/u9Ls2LFDABDfffed1fLMzEzh4eEhxo0bJy8rzzFer9eLiIgIMWLECLFgwQLx5ZdfCiHs/8y0Z/8v3EbF7VOlsTvApKSkCADi4YcfLnW9wgNmWlqaEEKIoUOHisDAQFFQUCCvc+3aNaFWq8XUqVPlZV27dhXNmzcXOTk58jKLxSI6dOgg6tevLy8rDCmdOnWyesyitxX9A2RlZdnUOGbMGOHu7m71XDExMQKAmDFjhrwsNzdXtGrVSgQGBlp9wBcHgAAg/vzzT3nZhQsXhMFgEAMHDpSXPfHEEyIoKMhqpxVCiCFDhghvb2+53sIAcufBxmw2C5PJJCZOnChvoxo1aohBgwYJjUYj0tPThRBCfPjhh0KtVstv8p07dwoAYtmyZVaPt2HDBqvl6enpwsfHRzz11FNW612/fl14e3tbLR8xYoQAYPV3FEKI++67T0RFRZW6vYSQPvgAiB07dsjLEhIShF6vF//+97/lZW+99Zbw8PAQp06dsrr/K6+8IjQajbh48aIQQohVq1YJAGLmzJlW2ys2NtZmWxa3X3zzzTc29UyfPt1mnypaf9EA8+qrrwqtViuSk5PlZbm5ucLHx0eMHj1aXmbvPlASAEKtVoujR49aLbf3cR9++GGrLw/FudsAI4QQK1asEADEtm3brNb74YcfBACxd+/eUp/7TgkJCUKn04kePXoIs9ksL58zZ44AID777DOb+go/9EtTuP++++678rJbt24Jo9EoVCqV+Pbbb+XlJ06cEADEpEmT5GU5OTlW9QghfWDr9Xqr94Q927to3cePHxfBwcGiTZs2VvuSve+D1atXCwDiv//9r7xOQUGBiI6OtivAtGzZUvTt27fUdcobYIxGo7h8+bK8/PfffxcAxL/+9S+bbVAoPj5eaDQa8c4771g9x19//SXc3Nzk5QUFBaJevXoiNDTUKtQIIeQvI0IIMW7cuGJDwAsvvCBMJpPN54k97nyv5uXliWbNmonY2Fir5QCETqcTZ86ckZcdOnRIABCzZ8+Wlw0YMEAYDAZx4cIFedmxY8eERqMpM8BYLBYREhIiHn30Uavl3333ndVx7W6O8a+88orN89n7mWnP/i/E3QUYu08hpaenAwC8vLxKXa/w9rS0NABAXFwcEhIS5CZkQDq1ZLFYEBcXB0Bqiv35558xePBgpKenIykpCUlJSbh58yZ69uyJ06dP48qVK1bP89RTT0Gj0ZRZt9FotHoNSUlJiI6ORlZWFk6cOGG1rpubG8aMGSP/rtPpMGbMGCQkJGDfvn1lPlf79u0RFRUl/163bl08/PDD2LhxI8xmM4QQWLVqFfr37w8hhPw6k5KS0LNnT6SmpmL//v2lPodarUaHDh3kU1PHjx/HzZs38corr0AIgT179gAAdu7ciWbNmsl9AVasWAFvb290797d6nmjoqLg6emJbdu2AQA2b96MlJQUDB061Go9jUaDBx54QF6vqGeeecbq9+joaJw7d67M7QUATZo0QXR0tPx7QEAAGjZsaHX/FStWIDo6Gr6+vlY1devWDWazWd4WGzZsgFarxVNPPWW1vcaNG2fzvEX3i5ycHCQlJaFdu3YAUObfoCRxcXHIz8/H999/Ly/btGkTUlJS5H29IvYBAIiJiUGTJk3k38vzuD4+Prh8+TL27t17V6/zbhXui2vXrkV+fr7d99uyZQvy8vLw4osvQq2+fch66qmnYDKZ8NNPP91TXU8++aRVjQ0bNoSHh4dVB+CGDRvCx8fHar/U6/VyPWazGTdv3oSnpycaNmxo9Tcsz/Y+cuQIYmJiEBYWhi1btsDX11e+zd73wbp16+Dm5oZnn31Wvq9Go8Hzzz9v1/bw8fHB0aNHcfr0abvWt8eAAQMQEhIi/962bVs88MADWLduXYn3+f7772GxWDB48GCr11urVi3Ur19fPhYdOHAA58+fx4svvmjT98meocc+Pj7IzMy0OkVmr6LHkVu3biE1NRXR0dHFvoe7deuGiIgI+fcWLVrAZDLJ+5TZbMbGjRsxYMAA1K1bV16vcePG6NmzZ5m1qFQqDBo0COvWrUNGRoa8fPny5QgJCUGnTp0A3N0xvui+VJQ9n5n27v9CCKucYA+7A0xhMCkMMiW5M+j06tUL3t7eWL58ubzO8uXL0apVKzRo0AAAcObMGQgh8MYbbyAgIMDqp/CcdmHnvEL16tWzq+6jR49i4MCB8Pb2hslkQkBAAP75z38CAFJTU63WDQ4OtumcVFijPSMq6tevb7OsQYMGyMrKQmJiIhITE5GSkiL3ISr6M2rUqGJfZ3Gio6Oxb98+ZGdnY+fOnQgKCkLr1q3RsmVL7Ny5E4B0HrVoMDh9+jRSU1MRGBho89wZGRny8xYetGJjY23W27Rpk019BoPBpj+Ur69vsf1qilP0jVrS/U+fPo0NGzbY1NOtWzcAt7fZhQsXEBQUBHd3d6vHi4yMtHmO5ORkvPDCC6hZsyaMRiMCAgLkferO/cJeLVu2RKNGjWz2dX9/f8TGxgJAhe0Dd+7/5Xncl19+GZ6enmjbti3q16+PcePGYffu3Xf1mssjJiYGjz76KKZMmQJ/f388/PDDWLJkCXJzc0u934ULFwBIIaIonU6H8PBw+fa7Udz+6+3tjdq1a9t8+Hl7e1vtlxaLBR999BHq168PvV4Pf39/BAQE4PDhw1b7UHm2d//+/eHl5YWNGzfCZDJZ3Vbe94Gnp6fV/e/cfiWZOnUqUlJS0KBBAzRv3hwTJkzA4cOH7bpvSUo6NpZ2XD19+jSEEKhfv77Naz5+/Lj8es+ePQsAdz0nydixY9GgQQP07t0btWvXxujRo4vtm1KctWvXol27djAYDPDz80NAQADmz59f7DGkrGNdYmIisrOzi91W9v7t4uLikJ2dLfcdzcjIwLp16zBo0CB5fy7vMd7Nza3Efqb2fGZW5vHG7i7f3t7eCAoKKnNHPnz4MEJCQuQ3n16vx4ABA/DDDz9g3rx5uHHjBnbv3o13331Xvk9hB7SXXnqpxKR554dQ0eRbkpSUFMTExMBkMmHq1KmIiIiAwWDA/v378fLLL9t0fKtshc/3z3/+EyNGjCh2nRYtWpT5OJ06dUJ+fj727NmDnTt3ykElOjoaO3fuxIkTJ5CYmGgVYCwWCwIDA7Fs2bJiH7PwIF5Y49KlS1GrVi2b9e4cJWBPK1hpSrq/1KIIuabu3btj4sSJxa5b+IYpj8GDB+PXX3/FhAkT0KpVK3h6esJisaBXr173tF/ExcXhnXfeQVJSEry8vPC///0PQ4cOlbdbRe0Dd+7/5Xncxo0b4+TJk1i7di02bNiAVatWYd68eXjzzTflYaYlfXO1p3N2SVQqFVauXInffvsNa9aswcaNGzF69GjMmDEDv/32m80HblUoaf+zZ79899138cYbb2D06NF466234OfnB7VajRdffNFqH7Jnexd69NFH8cUXX2DZsmVW32yBynkfFKdz5844e/YsfvzxR2zatAmffPIJPvroIyxYsEBurVKpVFbbotC97B93slgsUKlUWL9+fbF/j4raXwIDA3Hw4EFs3LgR69evx/r167FkyRI8/vjjxXbKL7Rz50489NBD6Ny5M+bNm4egoCBotVosWbKk2A7a9uxT96pdu3YICwvDd999h2HDhmHNmjXIzs6WW4CB8h/ji7Y03o3y7P/lVa4xa/369cPixYuxa9cuuTmqqJ07dyI+Pt7mjRcXF4cvvvgCW7duxfHjxyGEsNqghUOutVqt/G2iImzfvh03b97E999/j86dO8vLz58/X+z6V69etRkidurUKQAosWd9UcU1uZ46dQru7u5yQPDy8oLZbC7zdZbW9Nm2bVvodDrs3LkTO3fulEfKdO7cGYsXL8bWrVvl3wtFRERgy5Yt6NixY6nhr7CJMzAwsEL/FvciIiICGRkZZdYTGhqKbdu2ISsry6oV5syZM1br3bp1C1u3bsWUKVPw5ptvysuL+/uVd/bLuLg4TJkyBatWrULNmjWRlpaGIUOGyLcHBATYvQ+UR3kf18PDA3FxcYiLi0NeXh4eeeQRvPPOO3j11VdhMBjg6+tb7KgVe1o7ytpm7dq1Q7t27fDOO+/g66+/xj/+8Q98++23VqdyigoNDQUgzbtTdHqGvLw8nD9/XrH9dOXKlXjwwQfx6aefWi1PSUmBv7+/1bKytneh6dOnw83NDWPHjoWXlxeGDRsm31ae98HWrVuRkZFh9SF/8uRJu1+bn58fRo0ahVGjRiEjIwOdO3fG5MmT5b+Rr69vsaeJS9o/Sjo2lnZcjYiIgBAC9erVKzWcFR6zjhw5Uuq2KW2/1Ol06N+/P/r37w+LxYKxY8di4cKFeOONN4ptwQWAVatWwWAwYOPGjVZTKixZsqTE5ylNQEAAjEZjsduqPH+7wYMHY9asWUhLS8Py5csRFhYmnx4HKvYYb+9npr37f3mVK1ZNmDABRqMRY8aMsRpuCkhN8s888wzc3d3lD9RC3bp1g5+fH5YvX47ly5ejbdu2Vk3ggYGB6NKlCxYuXIhr167ZPG9iYmJ5ypQVJt6iCTcvLw/z5s0rdv2CggKrIWt5eXlYuHAhAgICrPq2lGTPnj1W5z4vXbqEH3/8ET169IBGo4FGo8Gjjz6KVatW2QwhBKxfZ+EOUdyHiMFgQJs2bfDNN9/g4sWLVi0w2dnZ+PjjjxEREYGgoCD5PoMHD4bZbMZbb71V7OsufJ6ePXvCZDLh3XffLbafwt3+Le7F4MGDsWfPHmzcuNHmtpSUFBQUFACQas/Pz7eaXM5iscjDJwsVt18AwMyZM20ev7S/Q3EaN26M5s2by/t6UFCQVZAszz5QHuV53DvfuzqdDk2aNIEQQv6bR0REIDU11arF9dq1a3ZNVFnSNrt165bNNm/VqhUAlHoaqVu3btDpdPj444+t7v/pp58iNTUVffv2LbOmyqDRaGxez4oVK2z669mzvQupVCosWrQIjz32GEaMGGE1jYS974M+ffqgoKAA8+fPl283m82YPXu2Xa/rzno9PT0RGRlp9TeKiIiQW3oLHTp0qMRTA6tXr7baLn/88Qd+//139O7du8Q6HnnkEWg0GkyZMsVmOwsh5Dpbt26NevXqYebMmTb7XNH7lbRf3vl61Wq13FpZ2n6p0WigUqmsWp3i4+OxevXqEu9TGo1Gg549e2L16tVWw4mPHz9e7N+8JHFxccjNzcUXX3yBDRs22EzmWJHHeHs+M+3d/+9mGHW5WmDq16+PL774Av/4xz/QvHlzPPHEE6hXrx7i4+Px6aefIikpCd98841VRyVAall55JFH8O233yIzM7PY62LMnTsXnTp1QvPmzfHUU08hPDwcN27cwJ49e3D58uVi51YoS4cOHeDr64sRI0Zg/PjxUKlUWLp0aYlNdsHBwZg2bRri4+PRoEEDLF++HAcPHsSiRYug1WrLfL5mzZqhZ8+eGD9+PPR6vRyUijaTvf/++9i2bRseeOABPPXUU2jSpAmSk5Oxf/9+bNmyBcnJyQCkA4SPjw8WLFgALy8veHh44IEHHpCDX3R0NN5//314e3ujefPmAKQg2LBhQ5w8edLm2jQxMTEYM2YM3nvvPRw8eBA9evSAVqvF6dOnsWLFCsyaNQuPPfYYTCYT5s+fj+HDh6N169YYMmQIAgICcPHiRfz000/o2LEj5syZU+6/xb2YMGEC/ve//6Ffv34YOXIkoqKikJmZib/++gsrV65EfHw8/P39MWDAALRt2xb//ve/cebMGTRq1Aj/+9//5G1a+A3MZDKhc+fO+O9//4v8/HyEhIRg06ZNxbbMFb4JX3/9dQwZMgRarRb9+/cvdSKnuLg4vPnmmzAYDHjiiSdsml/t3QfKy97H7dGjB2rVqoWOHTuiZs2aOH78OObMmYO+ffvKfdeGDBmCl19+GQMHDsT48eORlZWF+fPno0GDBmV2Mm7VqhU0Gg2mTZuG1NRU6PV6xMbG4uuvv8a8efMwcOBAREREID09HYsXL4bJZEKfPn1KfLyAgAC8+uqrmDJlCnr16oWHHnoIJ0+exLx589CmTRu5T1tV69evH6ZOnYpRo0ahQ4cO+Ouvv7Bs2TKbSTzt2d5FqdVqfPXVVxgwYAAGDx6MdevWITY21u73Qf/+/dGxY0e88soriI+PR5MmTfD999/b3berSZMm6NKlC6KiouDn54c///wTK1eutLrW2+jRo/Hhhx+iZ8+eeOKJJ5CQkIAFCxagadOm8gCOoiIjI9GpUyc8++yzyM3NxcyZM1GjRo0ST4cB0jHw7bffxquvvor4+HgMGDAAXl5eOH/+PH744Qc8/fTTeOmll6BWqzF//nz0798frVq1wqhRoxAUFIQTJ07g6NGj8od/4Xt5/Pjx6NmzJzQaDYYMGYInn3wSycnJiI2NRe3atXHhwgXMnj0brVq1QuPGjUusr2/fvvjwww/Rq1cvDBs2DAkJCZg7dy4iIyPvus/QlClTsGHDBkRHR2Ps2LEoKCiQ51Cx9zFbt26NyMhIvP7668jNzbU62wGgQo/x9nxm2rv/N27cGDExMeXryFuuMUt/O3z4sBg6dKgICgoSWq1W1KpVSwwdOtRqvpM7bd68WQAQKpVKXLp0qdh1zp49Kx5//HFRq1YtodVqRUhIiOjXr59YuXKlvE7hUOnihmIWN4x69+7dol27dsJoNIrg4GAxceJEsXHjRpthnjExMaJp06bizz//FO3btxcGg0GEhoaKOXPm2LVNAIhx48aJr776StSvX1/o9Xpx33332QwlFUKIGzduiHHjxok6derI269r165i0aJFVuv9+OOPokmTJsLNzc1m+ONPP/0kAIjevXtb3efJJ58sde6aRYsWiaioKGE0GoWXl5do3ry5mDhxorh69arVetu2bRM9e/YU3t7ewmAwiIiICDFy5EirYeIjRowQHh4eNs9R3FDb4oSGhhY7XLO4IZrp6eni1VdfFZGRkUKn0wl/f3/RoUMH8cEHH1gNcU9MTBTDhg0TXl5ewtvbW4wcOVLs3r1bALAaFnv58mUxcOBA4ePjI7y9vcWgQYPE1atXbYbKCiENXw0JCRFqtdpq/7pz+HCh06dPy8Pqd+3aVexrt3cfKE7hvna3j7tw4ULRuXNnUaNGDXmOhwkTJojU1FSrx9q0aZNo1qyZ0Ol0omHDhuKrr76yaxi1EEIsXrxYhIeHy8M/t23bJvbv3y+GDh0q6tatK/R6vQgMDBT9+vWz2qdKM2fOHNGoUSOh1WpFzZo1xbPPPmszbLa8w6iL238LjwV3unN/zcnJEf/+979FUFCQMBqNomPHjmLPnj02+68927u4urOyskRMTIzw9PQUv/32mxDC/vfBzZs3xfDhw4XJZBLe3t5i+PDh4sCBA3YNo3777bdF27ZthY+PjzAajaJRo0binXfesZlK4quvvhLh4eFCp9OJVq1aiY0bN5Y4jHr69OlixowZok6dOkKv14vo6GibOXpKOm6sWrVKdOrUSXh4eAgPDw/RqFEjMW7cOHHy5Emr9Xbt2iW6d+8uvLy8hIeHh2jRooXVEOWCggLx/PPPi4CAAKFSqeTnWrlypejRo4cIDAwUOp1O1K1bV4wZM0Zcu3at1O0khBCffvqpfLxv1KiRWLJkSbGvo6T3bHHvnV9++UVERUUJnU4nwsPDxYIFC+w+phZ6/fXXBQARGRlZ4jr3cowXwv7PTHuPN7iLYdSqv+/o8rp06YKkpKRim9/Jua1evRoDBw7Erl270LFjR6XLIaJivPHGG3jvvffkU2FEZbn7rsVEDujOi8EVnvs3mUxo3bq1QlURUVmuXbtm0/mZqDS8chZVK88//zyys7PRvn175Obm4vvvv8evv/6Kd999166h90RUtc6dO4cffvgBK1asQL9+/ZQuh5wIAwxVK7GxsZgxYwbWrl2LnJwcREZGYvbs2VYdEInIcezYsQNTpkxBly5d8OGHHypdDjkR9oEhIiIip8M+MEREROR0GGCIiIjI6TDAEBERkdNhgCEiIiKnwwBDRERETocBhoiIiJwOAwwRERE5HQYYIiIicjoMMEREROR0GGCIiIjI6TDAEBERkdNhgCEiIiKnwwBDRERETocBhoiIiJwOAwwRERE5HQYYIiIicjoMMEREROR0GGCIiIjI6TDAEBERkdNhgCEiIiKnwwBDRERETocBhoiIiJwOAwwRERE5HQYYIiIicjoMMEREROR0GGCIiIjI6TDAEBERkdNhgCEiIiKnwwBDRERETocBhoiIiJwOAwwRERE5HQYYIiIicjpuShdAZJesLCAzE8jLA/LzS/8RAlCrAZVK+letBtzcpB+tVvpXpwPc3QEPD8DTU1pOREROgwGGHENeHpCWBqSnW/+kpQEZGUBBQeU+v053O8wU/ddkAmrUkG4nIiKHoRJCCKWLIBeTlwckJEg/iYnST1aW0lWVzmQC/P2tfwwGpasiuk1YgIIswJwNWPIBYQZEAWApkP4PIf2Iv/9VaQC17u8fbZH///27SqXwCyIqHQMMVS6LBUhKuh1WEhKA1FSlq6oYHh5SkAkIAIKDgcBA6XQVUWXITwfyUoCCTMCcJf1b9P/mHEghpYKotIDGAGhNxf+oedqVlMUAQxUvJQW4dEn6uXYNMJuVrqhq6HRASAhQuzZQp450GoqovCxmIC8ZyL0J5N38+99kwJKndGXW1H+HG503oA8ADIGA3h9QMcRT1WCAoXtnsQBXrwLx8VJoSU9XuiLH4ONzO8wEBUmdh4nulHcLyLoC5NyQwkp+Kiq0JaUqqTSArgZgqCkFGkNNQMsgT5WDAYbuTkEBcPEicP68FFryHOzboaPRaKQgExkJ1K3LMOPKCrKkwJJ9WfrX7OD9v+6Vxl0KM+61Afe6DDRUYRhgqHwSE4ETJ4CzZxla7pZWC4SFAfXrS6ec2FmyerMUANnXgKzLQPYV6XSQK9P6AB51AY9QwFCL+z/dNQYYKltuLnD6tBRckl384FvRPD2lINOwoTTSiaoHixnIugRknAEyL0qjgciW2iAFGc96gDEEUGuUroicCAMMlezyZeDkSalvi6t0xFVSUBDQtClQrx6/lTojIaQWlvTTQEY8IPKVrsi5qHWAZzhgaiSdciIqAwMMWSsoAI4fB44cYWdcpZhMQIsWUquMht9IHV5eCpB2Sgou5kylq6kedH6AqTHgVR/QcBJJKh4DDEny8oBjx4DDh4GcHKWrIQAwGqUWmaZNAb1e6WroTpkXgFuHgZxrSldSfancpNNLpkaAMUjpasjBMMC4upwcqbXlyBF2ynVUbm5Ao0ZSqwznllGWpQBIPwWk/PX3cGeqMlofKch4N+YkegSAAcZ1ZWVJrS3Hj0sXQCTHp1YDERFAVBQ7/Fa1gmwg9SiQegywsIVSUWoD4NMM8G7G00sujgHG1eTmAvv2ScGFHXOdk1oNNGsGtG7Ni0xWtrwUIOWw1L9F8P3iUNQ6KcT4NAc0PMXqihhgXIUQUh+XP/+UQgw5P4NBao1p3JjXYKpo+RlA8l4puJBjU2kB76aAbwvp2k3kMhhgXMG1a8Du3ZzDpbry8QHatZNm+KV7Y84Dbh0AUo+wxcXZqNwA7yaAbysGGRfBAFOdZWQAv/0GnDundCVUFUJCgPbtAT8/pStxPsIi9XFJPsA+Ls5OrQf87pfCDOdTqtYYYKojsxk4dAg4eFCa14Vch0olDbtu25bXW7JX+lng5l6gIE3pSqgi6fyAgE6AsZbSlVAlYYCpbq5cAXbs4CR0rs5kArp0AWrx4F2inEQgcReQm6h0JVSZPCMB/3aAm7vSlVAFY4CpLgoKpNNFx44pXQk5CpVKGq3Upg1bY4qymIHkfUDKIQA8/LkElRbway2NWFKxw3t1wQBTHVy/DmzfDqSxCZyK4eMjtcYE8voyyL4BJPwC5KcoXQkpQesNBEQD7sFKV0IVgAHGmVks0rDoQ4ekYdJEJVGppJl877/fNa+vZCkAbv4hddRlqwt5NwNqPMCrXzs5BhhnlZYG/PwzkJCgdCXkTHx9gdhYoEYNpSupOllXgYQd7KRL1rQ+QM0HAUOA0pXQXWKAcUanTwO7dvESAHR3NBqgUyfpatfVmaUASPoNSGO/MCqJWuob43sfh1w7IQYYZ2KxAHv2AEePKl0JVQcNG0pBpjqeUspLAa5vAfI4eSPZwRgstca4eShdCZUDA4yzyM0FtmyRhkkTVRR/f6BHj+p1lev0M0DCTkCwhZLKQW0AAjsDnmFKV0J2YoBxBikpwIYNHGVElcNgALp3B4KClK7k3ggLkPgrTxnRvfFtDdS4X+kqyA4MMI7u0iVg61YgL0/pSqg6U6uBjh2lC0M6o4Is4PpmIOeG0pVQdeBRTzqlpOb8SY6MAcaRHT4M/P47h0hT1WnaFOjQwbk6NGbfkMKLOUvpSqg60fsDQT3ZL8aBMcA4IrMZ2LkTOHVK6UrIFYWHS0Ot1U4wY2naKWmINCxKV0LVkcYdCOoBGDgJpCNigHE0BQXAxo3srEvKqlNH6hfjyJcguHUIuPm70lVQdafSAIExgFek0pXQHRhgHEl+vtRZ99o1pSshki4E2asXoNMpXYmtpN+AlMNKV0GuxKcV4N9W6SqoCAYYR5GXB6xfD9xgJ0RyIP7+QJ8+0kglRyCEdC2jdJ5eJQV4NZBaY5ypj1g1xgDjCHJzgZ9+ApKSlK6EyJaPjxRilJ4rxlIAXN8KZF1Qtg5ybZ6R0gglhhjFMcAoLScHWLsWSOaMoeTAPD2Bvn0Bb29lnt+cB1zbAORcV+b5iYryDAdqxgIqJ+joXo0xwCgpK0sKLykpSldCVDajEXj4YcBkqtrnLcgCrq7jZQHIsXiEAbW6McQoiFteKVlZwJo1DC/kPLKzgXXrpH23qphzGV7IMWXGA9c2ARaz0pW4LAYYJeTlSR8EqalKV0JUPmlpUmfzqpgZ2lIAXF3P8EKOK+sicJ0hRikMMFXNYpEuysg+L+Ssbt4ENm2SJlysLBYzcG0jkJtQec9BVBGyLkn9swQnU6xqDDBVbccO4PJlpasgujdXrwI//1w5l7kQArixFcjmZI7kJLKvADe2K12Fy2GAqUp//snLA1D1cf48sGtXxT9uwg6pfwGRM8k4AyT9oXQVLoUBpqqcOAHs3690FUQV6/hxKZhXlMQ9QPrJins8oqqUchBIPaF0FS6DAaYqXLpUOd9UiRzB/v3AyQoIHckHgNS/7v1xiJSUuAvIvKR0FS6BAaayJSVJnXYt7OBF1diuXfc2k3RGPJC8t8LKIVKOBbi+BcjhzOqVjQGmMuXkSKM18vOVroSocpnNwObN0mUxyisvBUjYXtEVESlH5Esjk/IzlK6kWmOAqUzbtgEZ3IHJRaSnS/t8eVjy/54MrArmlSGqSuYsKcRY+AW2sjDAVJb9+6W+L0Su5OLF8nVWv7EdyE+prGqIlJWXLF09nSoFA0xluHIF2LdP6SqIlLFvn/QeKMutg0Dm+Uovh0hRGeeAlKNKV1EtMcBUtOxsqRmd18gkVyUEsHVr6adPsy4DN9lpl1xE0h4gh7NKVzQGmIq2fXvVXuyOyBHl5JQ8+i4/Hbi+FQBDPrmKv0cmmdnXqyIxwFSkw4fZ74WoUEKC7alUYZHCi+UuRisRObOCDCBxh9JVVCsMMBXl5k3gD04jTWTl4EEgMfH277cO8gKN5LoyzgFpnGm6ojDAVAQhgF9+4WR1RHcSQjqtajYDOYlAMi+nQS4ucTeQl6p0FdUCA0xFOHLk3mYhJarObt0C/vwDuLENAEM+uThRIL0XONDjnjHA3KuMDGAvR1MQlSrjLJDLzu1EAKTTqKnHlK7C6THA3Kvdu4GCAqWrIHJcPjrAIwtIBwceERVK3gsUMNTfCwaYe3H+PHDhgtJVEDkuFYDaf/8/Pw/Id1eyGiLHYcmT+sPQXWOAuVt5eVLrCxGVLNQdUBeZ+yI9G7C4KVcPkSPJPA9kXlS6CqfFAHO39u7lhHVEpTG6AaZs62VCANlaZeohckSJuwALuyHcDQaYu5GQABxjByyiUtXVothOL9nZgFlX5eUQOaSCDCD5T6WrcEoMMHfjt984BI6oND46QJ9d8u1ZqqqrhcjRpRwBcm8qXYXTYYApr0uXgOvXla6CyLGFlHF7bi5QYKiSUogcn4Udeu8CA0x5cc4XotIFGQGNHRetyzBzWDVRoZzr0lXayW4MMOVx7hxn3CUqjUYF+Ofbt25BPpBvrNx6iJzJTX5BLg8GGHsJAfzJjlZEpapjAFTlGFGRkQcIHoaIAAC5iUBGvNJVOA0eOex1+jSQkqJ0FUSOS6sGTLnlu4/FDOSxLwyRLPlPDhKxEwOMPSwWYN8+pasgcmy1DbirizVm5gCCo5KIAAB5ydK1w6hMDDD2OHECSE9Xugoix+WmAjxz7u6+Fgv7whAVdfNPQPDK7WVhgCmLxQIcOKB0FUSOLcQIqO7hgJuZxxFJRIUK0oC0U0pX4fAYYMpy/jyQmal0FUSOS4Xy9325k7mA88IQFXVrP1thysAAU5YjR5SugMixBRsBlfneHyeLB2siWUEGkBmvdBUOjQGmNElJwI0bSldB5Nh87Zz3pSz5eYCZrTBEspSjSlfg0BhgSnOUOw9RqYKMgLoCr6SbxY4wRLKca0BustJVOCwGmJLk5ABnzihdBZFj86uAU0dF5eYCFm3FPiaRM0vlF+mSMMCU5ORJwFzBB2ei6sRLa981j8orjwGGSJZ+GjBXwvusGmCAKY4QPH1EVJbASgoaWbkcUk1USBQA6SeVrsIhMcAU58IFICND6SqIHJcKgPs9Dp0uicXMzrxERaUeU7oCh8QAU5wTJ5SugMixBRoqZuh0SXLYBEMky08FMi8pXYXDYYC5U04OcPmy0lUQOTa/Sn78nFxAaCr5SYicSBq/WN+JAeZO585Jlw8gouIZNID2Lq97VB75+sp/DiJnkXWRnXnvwABzp7O8CihRqWpVUbDIrsD5ZYicnTBzZt47MMAUlZkJXL+udBVEjs2zgmbeLUt+HueEISoqg1+wi2KAKer8eWkINREVz8MNUFdRgAGAfAYYIlnWFcBcBadvnQQDTFHx8UpXQOTYalRxoMjlaSSi2yxA5gWli3AYDDCFcnKAa9eUroLIsXlWcQf3vDzAwsMUkSwjXukKHAaPDIUuXuTpI6LS6NSAWyVNXlcaM0cjEcmyLwOWKjyN68AYYArx9BFR6WrolHleHquJbhNmIIuT2gEMMLfx9BFR6UwKPW82r41EZIWz8gJggJEkJwO5CjSNEzkLtQrQKzSJlrAAFp5GIpJl8ws3wAAjYesLUen89AAUnKE6n5cVIJIVpAEFmUpXoTgGGIABhqgsXgofKvI4nJrIClthGGAAMMAQlcWo8PXB8vIAoVK2BiJHwgDDAIOUFCA7W+kqiBybmwNcRM6i0CgoIkfEAMMAw9YXojJ4aaFo/5dCZvaDIZLlp7j8ZQUYYBhgiErn7aZ0BZICjqUmsuLirTAMMAwwRKXzULqAv+VyRjsiKwwwLiwnB8jkUDSiUmkdJDiYCwALTyMRyXJuKF2Bolw7wKSkKF0BkWPTawC1Aw1hZkdeotvybildgaJcO8Dccu0/PlGZvLRKV2CtgEOpiWSiAMjPULoKxbh2gGELDFHpjA4WGCzsyEtkxYVbYVw7wLAFhqh0egcLMAVmpSsgciz5KUpXoBjXDjBsgSEqndYB5n8pKt+B+uMQOQK2wLig/Hwgw3XPHRLZxc3BAoOwcCQSUVF5KUpXoBjXDTBsfSEqnVoFqBwswACAcJCJ9YgcAQOMC2L/F6LSeThoULC47mGLyIYlx2UvKeC6R4K0NKUrIHJs7g56qsbsYB2LiZTmoq0wrhtgclwzsRLZzeCghwezg3UsJlKaOUvpChThoEeoKpCdrXQFRI7NQc8gcS4YojvwFJKLYQsMUek0DhoULGyBIbLCAONiGGCISqd20ADDU0hE1syueUbBdQMMTyERlc5hW2A4Gy+RFbbAuJjcXKUrIHJsKgdu6RCue+gissEA40JycgDhoN8uiRwFAwyRc2CAcSHs/0JUNpUDn6phgCG6zUX7wDjqQMnKxQBT4XacOoXpmzZh38WLuJaaih+efRYDWrWSb/9+/34s2LED+y5eRHJmJg785z9oVaeO1WPk5Ofj3ytW4Ns//0RuQQF6NmmCecOGoabJJK+jGjPG5rm/efJJDGnTptJem0vSqgGUvwUmPTMHbyz6ET/sOIiE5HTc16AOZv0rDm2ahAEAJn+yBt9u3otLCbeg07ohqmFdvPPMADzQtF75nogBpsLs+OMUpn+yCfuOXsS1hFT8MO9ZDOjeSr5dCIFJs9Zg8Xc7kZKWjY5REZg/ZRjqh9W0epyftv2FqXPW4vDJKzDotYhpWx+r54+Vbx8/9Vvs3n8WR05dReOIWji45o2qeonVH1tgXEiBA17fxcll5uWhZe3amDt0aIm3d4qMxLRHHinxMf713XdYc/gwVjz9NH75979xNSUFjyxYYLPekhEjcO2//5V/igYlqiDauzs0PPnel9i89ziWvjkKf331Jno80ATdxn+EKwnSpTsa1KmJOf8eir++ehO7FkxAWFAN9HhhJhJvpZfzmTgbb0XJzM5Dy0a1MXdS8e/d/y7aiI+//BkLpv4Dv698BR5GPXqO+hg5ufnyOqs27MfwCZ9h1KMdcGjNG9j97QQM69/W5rFGP9YBcX3vr7TX4ros0oVOXYxrtsBwHokK17tZM/Ru1qzE24e3awcAiE9KKvb21OxsfLp7N75+4gnENmoEAFgyciQaT5qE386dQ7vwcHldH3d31PL2rsDqycZd5IPsnDys2n4AP04bi873NQAATH6yP9bsOoz5P/yCt8cMwLCe1h9qH74wCJ+u2Y3DZy6ja5vG9j8Zu7BVmN4xzdA7pvj3rhACM7/Yiv+M7YOHu7UCAHw5fRRqtnsJqzcfxJB+bVBQYMYLby/H9JcfxRODOsn3bVI/2OqxPn5zCAAgMXkNDp+4XDkvxpUJ4XK53jVbYNiB1+Hsu3AB+WYzujW+/SHWqFYt1PXzw55z56zWHffNN/D/v/9D2/few2e7d0Pw71nxVOU/EhaYLTCbLTDorL8XGfVa7Dp01mb9vPwCLFq9E96eRrSsX8fm9tLrK3d5dBfOX0rC9cQ0dOtw+33p7WXEAy3rYc8B6X25/+hFXLmRArVKhfseehtBHSag9xMf48ipK0qV7aJc74u5a7bA8APP4VxPS4POzQ0+7u5Wy2uaTLiemir/PvWhhxDbsCHcdTpsOnYMY7/+Ghm5uRgfG1vVJVdvdxEQvDwMaN8sHG8tWYfGYUGo6WfCN5v/wJ4j5xBZO1Beb+2uwxjy5ifIyslDUA1vbJ71Ivx9PMv3ZHwLV4nrSdJFb2v6m6yW1/Q34XqS9L48d0lqVZ08ey0+fHUQwmrXwIxPN6PLP2fg1Ka34OfjUbVFuyoX/FxjCww5lTf69kXHyEjcV7cuXu7VCxN79sT0TZuULqv6ucsWjqWTRkMIgZCHXoY+Zhw+/m4bhnZvA3WRFp0Hoxri4Bf/wa+LJqJXu6YY/J9FSEjm1eGdleXva1O9/mxvPNqrNaKahWLJ+yOgggor1u9TuDpX4nqfa64ZYMjh1DKZkFdQgJQs66uq3khLK7W/ywP16uHyrVvIzc8vcR2qOhG1A/DL/JeQ8fPHuLT6ffzx2avILzAjPMRfXsfDqEdknUC0axaOT19/HG4aDT5ds1vBqqkktf5uebmRZB0wbySloZa/9L4MCpT+bRJ5u8+LXq9FeB1/XLyWXEWVkityzQCjds2X7ciiQkOh1Wiw9cQJednJ69dxMTkZ7Yt04L3TwUuX4OvuDr1WWxVluo57PJ3uYdQjyN8bt9IysfH3Y3g4umXJTyUsyM3nyEBHVK+OP2oFmLB1z+33ZVp6Nn4/dB7t75Pel1FN60Kvc8PJ89fldfLzzYi/chOhwX5VXrPLUrne55pr9oG5iw6KVLqMnBycSUyUfz+flISDly7Bz8MDdf38kJyZiYvJybiakgJACieA1PJSy9sb3kYjnujYEf+3YgX8PDxgMhjw/Lffon14uDwCac2hQ7iRno529erBoNVi8/HjeHf9erzUvXuVv95q7y5bozf+dhRCCDQMrYUzlxMwYc4qNAqthVH9OiIzOxfvfL4OD0W3RFANbySlZmDuyu24kpiCQbFRFVs/2S0jMwdnLhR5715OwsFjl+Dn44G6wX54cURXvD1vHeqHBaJebX+8MfNHBAf6yHPFmLyMeGZoZ0yatQZ1avkhNMQP0z+RTusO6n3773rmQgIyMnNxPTEN2bn5OHjsEgCgSWQQdDrX/CiqWK73uaYSrjiE49IlYP16pauoVrafPIkHP/zQZvmI9u3x+ciR+PzXXzHqiy9sbp/Urx8m9+8P4PZEdt/s3Ws1kV3hKaQNR47g1dWrcSYhAQJAZEAAno2JwVOdOkHNVrWK5eEGRJS/VeS7LX/i1QU/4HJCCvxM7ni0S2u888wAeHsakZObj2GTPsHvR+ORlJqBGt4eaNM4DP8Z2Uee6M5uPkbAzTVnH61o238/iQf/Wcx7d2B7fP7fkfJEdouW70RKWhY63R+JeZOHoUG92xPZ5eeb8eqMH7B09W/IzsnHAy3rYeZ/BqNpkaHUXf4xA7/8ccrmec5vewdhtf1tllM5RTzpcq0wrhlgrl4F1q5Vugoix2XQAA0c+FICPgbAzTVnHyUqVsRTLnd2wbXiWiG9XukKiBxbrgOHFwBQud73LqISqfUuF14AVw0wBoPSFRA5NgE49OHBka+UTVTVNK75mebAR6hKxABDVDZHvmAiAwzRbRqj0hUowoGPUJVIowE47JaodBZHPjw4+CkuoqrEFhgXw1YYotI5aoBRqV1xxChRyRhgXIzRNZvciOxmcdCUoHHdwxZRsdQMMK6FLTBEpXPUszQuNtcFUZnYAuNiGGCISmdmCwyRU2CAcTEMMESly3PQuVY0DhqsiJTCUUguxtNT6QqIHFuWg55DYoAhsqb1UroCRbhugPHxUboCIseW6aBXiFZzDhii29SA1qR0EYpggCGi4uVZHHMyO5WDBisiJei8XbZju2u+akA6hcTJ7IhKZ3FTugJbagc9tUWkBK2P0hUoxnUDDMBWGKKyFGiUrsCamxsv5EhUlM5X6QoU49oBxtd1//BEdsl1sA6zbg7YIkSkJJ2P0hUoxrUDDFtgiEqX42CtHRyBRGSNLTAuii0wRKXLcrAOs2yAISpCBWi9lS5CMa4dYNgCQ1S6tHw41GGCHXiJbnPzBNSum+od6MikAJMJ0DhYJ0UiR5PvIKP1VCpAnad0FUSOQ++vdAWKcu0Ao1IBgYFKV0Hk2HIcJOTrdAC7wBDdZgxSugJFuXaAAYAg194BiMqU7iCnbbQ8XBFZYYBxcQwwRKVLyVe6Aombg42IIlKSWgfo/JSuQlEMMDVrAmpuBqISFVgAswP0g1E7SJAicgSGWlI3CBfGT243NyAgQOkqiBxbnsIjHTRuHIFEVJQxWOkKFMcAA/A0ElFZMhV+fp0DtAARORJjLaUrUBwDDMAAQ1SWZIWHL+uUfXoih6LSuvwQaoABRlKL5xKJSpVjBswKpggN538hkhlqAip+fHMLAIBWC/gzzRKVKkuhfjBaLfu/EBXlHqJ0BQ6BAaZQnTpKV0Dk2JIVui6Sgf1fiKx4hCldgUNggCkUFqZ0BUSOLTUPEAq0wmgd7IKSRErS+QI6172AY1EMMIX8/QFPT6WrIHJsOVXcGqLR8PpHREWx9UXGAFNUvXpKV0Dk2FKqeDZcPYcfEVnx4OdUIQaYohhgiEp3MxdVetjQWqruuYgcnZsJMHDASSEGmKJq1gQ8PJSugshxWQSQq6+a51KpATeePiKSeUUoXYFDYYApSqUCwsOVroLIsd2solYRox5Q8QKORDJPBpiiGGDuFBmpdAVEji0pt2pGI+k59wuRTOcL6F376tN3YoC5U0AAYDIpXQWRY0uv5M61blrOvktUlFcDpStwOAwwxWnAHYWoVDcqOVwYOXkdkUzlBpgaKV2Fw2GAKU7jxoCam4aoRNkFQL6h8h5fm1t5j03kbDwjAE0VdZ53IvyULo7RyM68RGW5VUkXQNXree0joqK8mypdgUNigClJs2ZKV0Dk2G5kA0JT8Y9r5GGJSKYP5NwvJeCRoiSBgVKHXiIqngCQVcHN2moNoMmp2MckcmY+/DJdEgaY0jRlsx1Rqa7kAqjAU0kenPuFSKYxAp7szlASBpjSREQAhkrsqEjk7HLMQHYFvUdUakDH1hcimamx9L6gYnHLlEajARpx6BpRqa7mV8zjeBgAFa99RCRRA96NlS7CoTHAlKVJE+kSA0RUvMwCINd474+j48R1RDJTA8CN1+YrDQNMWTw9eXkBorJcK7i3+7sbAPU9PgZRtaEGfFsrXYTDY4CxR1QUJ7YjKk1a/r1NbGfgvC9EMu/GgNZT6SocHj+V7WEyAQ0bKl0FkWO7fpejhwwGQF1B/WiInJ3KDfC9T+kqnAIDjL1at5Y69RJR8W7l3l0rjJEdd4lk3k0BN3elq3AKDDD28vDgvDBEZblSzlNBRgOvOk1USK0DfFspXYXTYIApj1atAC2vkktUorR8IKscI5KM7LhLJPNpzos2lgMDTHkYDEDz5kpXQeTYLubCrkOLh5Ejj4gKqQ2ANz9fyoMBprxatJCulktExcuzACllvEfUasCQWzX1EDkDv9aARqd0FU6FAaa8dDrgPvYQJyrVpWzA4lby7Z6cdZdIpg+QOu9SuTDA3I1mzQA/P6WrIHJcAkBiCQHGzQ3QZlVpOUSOSwUERHPG97vAAHM31GogOlrpKogc240coKCYU0me2gq9gDWRU/NuChj8la7CKTHA3K2aNaXrJBFRyS5aYJVWDAbALVuxcogcisYDqNFG6SqcFgPMvWjbFnDnhENEJcrIB9L+HlatUgHuHHVEJAvoCKg5NcfdYoC5Fzod0KGD0lUQObYLWYBZB3hx2DSRzCMM8AxTugqnxgBzr8LDgbp1la6CyHEJAGkmQJejdCVEjkGlBfz55fdeMcBUhE6dpJEVRGRLrQYe6AL4RSldCZFj8H+AV5uuAAwwFcHTE7j/fqWrIHJMUVHStAO+rQBDTaWrIVKWRxjgzQEgFYEBpqI0bw4EBytdBZFjCQiQriEGSJ14az4IqNhaSS5K4w4Edla6imqDAaaiqFRAbKw0TJSIpNOqXbpYT9ClNfHcP7koFVAzFtDwM6KiMMBUJHd3ICZG6SqIHEN0NODra7vcuxFgalT19RApyfc+wJ2t9BWJAaaihYYCTXlNC3JxTZoA9euXfHtAJ0AfWHX1ECnJGMxO7JWAAaYytG8vnfsnckUBAWXPj6RSA0HdAY2xamoiUorGHajZldc6qgQMMJVBrQa6dwf0xVwHhqg6MxikfV9tx6HFzQOo1Q08DFH1pQZqdQXcGNQrA48clcXTE3jwQaWrIKo6hR3ZPcsxv4UxCPBvV3k1ESkpoJO0j1OlYICpTHXrAq1bK10FUdWIigJq1y7//XyaAV6l9Jchcka+90kd1qnSMMBUtvvvL70zI1F1cK9hPSAa0PtXXD1ESvKqz6tMVwEGmKoQE8NJ7qj68vOTTh3dC7UbENQbcDNVTE1ESjEGA4GcTqMqMMBUBbUa6NFDOtATVSeenkDv3tKV2e+VmxEI6QtoPO79sYiUoPMFavWQRtlRpeNWrio6nXSg9+DBmaoJgwHo06di92mtFxDSB1BztlJyMhp3qRVRUwFhnuzCAFOVPDwq7tsqkZLc3IBevQAfn4p/bJ0vENwbUGkr/rGJKoPKDQjqxStMVzEGmKrm52f/PBlEjqjwlGhgJc6kawgAgnoCKk3lPQdRRVBpgeA+gIGd0KsaP0WVEBLCayaR8+rS5e6GS5eXezAnuiPHptZJ4cVYS+lKXBKPDEqpX18KMZxempxJhw5AZGTVPZ9H6N8jOvg+IQej1gPBfQFjTaUrcVkqIYRQugiXduYMsG0bwD8DObo2bYD77lPmuTPigRtbAWFW5vmJilIbeNrIATDAOIJz54CffwYsFqUrISpex47KX2U96ypwfRNgyVO2DnJtagMQ0g/Qc1oMpTHAOIoLF4AtWwAzv2GSA1GrpVOdjjKbdE4ScG09YM5WuhJyRRp3aa4ina/SlRAYYBzL5cvApk1AQYHSlRABGg3QrRsQGqp0Jdby04ArPwEF6UpXQq7EzUs6baTzVroS+hsDjKO5ehXYsIEhhpSl1QI9ezruJTAKsoCr64C8ZKUrIVdgCAKCugMaTrDoSBhgHNGNG8DGjUBOjtKVkCvS66UJFytznpeKYM4Drm0Acq4rXQlVZ6ZGQEAnXh7AATHAOKr0dCnEJPMbJlUhd3egb1/A10nO8VvMQOIuIP2k0pVQtaMC/NsBPs2VLoRKwADjyPLzpdFJFy4oXQm5goAAaYZdZ7xeV8pRIGkPAI7kowqg1gE1uwIedZSuhErBAOPohAD27gUOHlS6EqrOIiOl0UYaJ566P/s6cH0zRyjRvXEzAcE9OdLICTDAOIszZ4BffuEwa6pYKpU0QV2rVkpXUjEKsoDrW9gvhu6OsTZQK5addZ0EA4wzSUyU+sVkZSldCVUHBgMQG1s11zWqSkIAN/cCKQeVroSchhqo0Qbwbal0IVQODDDOJisL2LxZGqlEdLcCA6U5Xjw9la6k8mReBG5sBywczUelcDMBtbpKV0Anp8IA44wsFmD/fuDAAV5DicqvSRPpooxqFxgWWpANJO0GMs4pXQk5Iq+GQEAHQK1VuhK6CwwwzuzaNelCkBkZSldCzsDdHejcGahbV+lKql7GeWm4NTv4EgBojEBgZ+lq5+S0GGCcXW4usGsXcPas0pWQI6tfX2p10euVrkQ55hwg8Vcg44zSlZCSPMOlienYUdfpMcBUF+fOSUGGs/dSUUYjEB0NhIUpXYnjyLwIJOwEzJlKV0JVyc0knS7ycMEWyGqKAaY6yc4Gdu4E4uOVroQcQUQE0LGjNNqIrJnzgJu/AWknlK6EKpvKDfC9D/BpAaideJ4jssEAUx2dOwf89hv7xrgqg0FqdalXT+lKHF9OApD0G+eNqa48w4Ea7QBtNR5t58IYYKqrggLg0CHph1e2dg0qFdCokTQxHVtdyicjHrj5O5CfqnQlVBF0voB/R8DdQa+mThWCAaa6y8gA9uwBzp9XuhKqTLVrA+3aAX5+SlfivIRFOqWUvI+jlZyVWg/4tQa8m/Lq0S6AAcZVXL0K/Porr25d3fj6SsGlDi86V2Es+cCtQ0DKYUCw9dIpqA2AbwspuHBOF5fBAONKLBbg2DFg3z5p+DU5L6MRiIoCGjeWTh1RxSvIApL3A+mnGGQclcZdmv7f1BhQuyldDVUxBhhXlJcHHDki/XDYtXPRaIDmzaWLL+p0SlfjGsw5QOoxIPUoTy05CjdPwKclYGrEkUUujAHGleXnA8ePA4cP8wKRjk6nky4B0KyZNKMuVT2LWZoEL+UwkHdL6Wpck5vp7xaXhuzjQgwwBMBsBk6ckEYscei1Y/HwkFpcGjcGtDy37zAyL0lBJvuK0pW4ADXgGSa1thhDeMqUZAwwdJvFApw+DRw8CKRyOKmi/PyAFi2AyEjXuOiis8pNBtKOA+lnAAv7lVUorY8UWkwNOO0/FYsBhmwJAVy4AJw8CVy8yCteV6WgIKBlS9e84KIzExYg8wKQdgrIugTAonRFzknlBnjWkzrlGmspXQ05OAYYKl1WFnDqlBRm2CpTOQwG6WKLDRtyHpfqwJwjXf064yyQfQ0AD7GlUwPuIYBHPWnmXA07p5N9GGDIfteuSUHm3DnO7nuv1Gpp8rmGDYHQUJ4mqq4KsqQwk3UJyL7K4diFVFrAvbbU2uJel6GF7goDDJVfXh5w9qzUX+bGDZ5iKo9ataR+LeHhnO7f1QgLkHMDyLoMZF0BchPhUq0zWh/pStDudaXTQxxFRPeIAYbuTU6OdPXr+HjgyhVpRBPd5uYGBAdLrS1hYYAnLypHfzPnSq0yWVekUFOQpnRFFUflBuj9AUMgYKgp/evmoXRVVM0wwFDFKSiQQsylS9JPerrSFSnDz08KLHXqSC0uGk60RXYw50ijmnKTgLzCf1PgFB2C3UzWYUVfgy0sVOkYYKjypKZKgebGDSAxEUhJUbqiymEwACEhUmipXVuau4WoIgiLNGle7k3pJz8VKMiUfixVPIu2xgPQmor/0eirthYiMMBQVcrLAxISpDCTkCD9ZDvZ1OweHkCNGoC//+0fnhYiJQiL1Em4IBMwZxb5f7Z0QUpRAAgzYPn7XwjpR/z9r0oDqHXSxQ/VumL+r5PmX9GaADcvTtlPDocBhpSVkSEFmZQU6ZRT4U9GhrKdgw0GKZiYTLeDSo0a0kUUiYhIcQww5JgsFinE3Blq8vNL/rGU0ldArZY61Gq1t/91d5dCioeH7b/st0JE5NAYYKj6MJulICOEFFhUKunfwh8iIqo2GGCIiIjI6fBrKRERETkdBhgiIiJyOgwwRERE5HQYYIiIiMjpMMAQERGR02GAISIiIqfDAENEREROhwGGiIiInA4DDBERVbnt27dDpVJh5cqVlfo8YWFhGDlyZKU+R3nEx8dDpVLhgw8+ULoUp8cAQ0TkYj7//HOoVCqoVCrs2rXL5nYhBOrUqQOVSoV+/fopUCFR2RhgKlF1/oYRFhZm14GtcBts37698osionIxGAz4+uuvbZb/8ssvuHz5MvR6vQJVEdmn2gYYfsNwDceOHcPkyZMRHx+vdClETqdPnz5YsWIFCgoKrJZ//fXXiIqKQq1atRSqrPqxWCzIyclRuoxqpdoGmEL8hqGszp07Izs7G507d66Uxz927BimTJnCAEN0F4YOHYqbN29i8+bN8rK8vDysXLkSw4YNK/Y+H3zwATp06IAaNWrAaDQiKiqq2FbmzZs3o1OnTvDx8YGnpycaNmyI1157rdR6cnNz0a9fP3h7e+PXX38FIH3wz5w5E02bNoXBYEDNmjUxZswY3Lp1y+q+Qgi8/fbbqF27Ntzd3fHggw/i6NGjZW6D/Px8+Pn5YdSoUTa3paWlwWAw4KWXXrKqcdKkSYiMjIRer0edOnUwceJE5ObmWt1XpVLhueeew7Jly9C0aVPo9Xps2LDBap2PPvoIoaGhMBqNiImJwZEjR8qsl26r9gGG3zCUpVarYTAYoFZX+12NyOmEhYWhffv2+Oabb+Rl69evR2pqKoYMGVLsfWbNmoX77rsPU6dOxbvvvgs3NzcMGjQIP/30k7zO0aNH0a9fP+Tm5mLq1KmYMWMGHnroIezevbvEWrKzs9G/f3/8+uuv2LJlCzp06AAAGDNmDCZMmICOHTti1qxZGDVqFJYtW4aePXsiPz9fvv+bb76JN954Ay1btsT06dMRHh6OHj16IDMzs9RtoNVqMXDgQKxevRp5eXlWt61evRq5ubnytrBYLHjooYfwwQcfoH///pg9ezYGDBiAjz76CHFxcTaP/fPPP+Nf//oX4uLiMGvWLISFhcm3ffnll/j4448xbtw4vPrqqzhy5AhiY2Nx48aNUuulIkQ1tWTJEgFArFixQqhUKrFu3Tr5ttzcXOHr6ytmzJghQkNDRd++fa3uO336dNG+fXvh5+cnDAaDaN26tVixYoXNc2zatEl07NhReHt7Cw8PD9GgQQPx6quvyrdv27ZNrqFQTk6O6Nu3rzCZTGL37t1CCCHMZrP46KOPRJMmTYRerxeBgYHi6aefFsnJyVbPZ7FYxFtvvSVCQkKE0WgUXbp0EUeOHBGhoaFixIgRZW6Tb775RrRu3Vp4enoKLy8v0axZMzFz5kz59kmTJonidonCbXn+/Hl5WeF227hxo2jZsqXQ6/WicePGYtWqVVb3LdwG27Zts1r+22+/iZ49ewqTySSMRqPo3Lmz2LVrl81zX758WYwePVoEBQUJnU4nwsLCxDPPPCNyc3Pluu78ufO5iMha4Xtn7969Ys6cOcLLy0tkZWUJIYQYNGiQePDBB4UQotjjY+F6hfLy8kSzZs1EbGysvOyjjz4SAERiYmKJNRQ9Pqanp4uYmBjh7+8vDhw4IK+zc+dOAUAsW7bM6r4bNmywWp6QkCB0Op3o27evsFgs8nqvvfaaAFDm8XHjxo0CgFizZo3V8j59+ojw8HD596VLlwq1Wi127txptd6CBQsEAPmYLoQQAIRarRZHjx61Wvf8+fMCgDAajeLy5cvy8t9//10AEP/6179KrZVuq/Zfi/kNQ7J582YMHToUvr6+mDZtGt5//3106dKl1HrLcvr0acTFxaF3795477335O1UtDm6OD///DM6d+6MtLQ0TJo0Ce+++y5SUlIQGxuLP/74Q17v6tWraNu2Lb799lvExcXh448/xvDhw/HLL78gKysLnTt3xvjx4wEAr732GpYuXYqlS5eicePGd/2aiFzN4MGDkZ2djbVr1yI9PR1r164t8fQRABiNRvn/t27dQmpqKqKjo7F//355uY+PDwDgxx9/hMViKfX5U1NT0aNHD5w4cQLbt29Hq1at5NtWrFgBb29vdO/eHUlJSfJPVFQUPD09sW3bNgDAli1bkJeXh+effx4qlUq+/4svvmjXNoiNjYW/vz+WL19u9do2b95s1bKyYsUKNG7cGI0aNbKqJzY2FgDkegrFxMSgSZMmxT7ngAEDEBISIv/etm1bPPDAA1i3bp1dNROqfwsMv2FIXnjhBWEymURBQUGJ65S3BQaAVYtLamqqCAoKEvfdd5/NNihsFbFYLKJ+/fqiZ8+eVq8jKytL1KtXT3Tv3l1e9vjjjwu1Wi327t1rU1PhfVesWMFWF6JyKnp8FEKIXr16iQEDBojPP/9c6HQ6cevWLSFE8cfHNWvWiAceeEDo9Xqrlk+VSiWvk5WVJTp27CgACH9/fxEXFyeWL18uzGazvE7hscHT01O4ubmJv/76y6bO3r17F9vKWvjz0EMPCSGEeO+99wQAcfbsWZvH8PX1tauFesyYMcLLy0vk5OQIIYT45JNPBABx8OBBeZ3GjRuXWs/48ePldQGI0aNH2zxPYQvMm2++aXPb8OHDhV6vL7NWklT7FhiA3zAK683MzCyzdaQ8goODMXDgQPl3k8mExx9/HAcOHMD169eLvc/Bgwdx+vRpDBs2DDdv3pRfb2ZmJrp27YodO3bAYrHAYrFg9erV6N+/P+6//36bxym6DYjo3gwbNgzr16/HggUL0Lt3b/n4dqedO3fioYcegsFgwLx587Bu3Tps3rwZw4YNgxBCXs9oNGLHjh3YsmULhg8fjsOHDyMuLg7du3eH2Wy2esyHH34YQgi8//77NsdSi8WCwMBAbN68udifqVOnVtg2GDJkCNLT07F+/XoAwHfffYdGjRqhZcuWVvU0b968xHrGjh1r9ZhFP0uo4rkpXUBVCAgIQLdu3fD1118jKysLZrMZjz32WInrr127Fm+//TYOHjxo1bO86IdmXFwcPvnkEzz55JN45ZVX0LVrVzzyyCN47LHHbDqsvvjii8jJycGBAwfQtGlTq9tOnz6N1NRUBAYGFltLQkICAODChQsAgPr169u8Nl9f3zK3wdixY/Hdd9+hd+/eCAkJQY8ePTB48GD06tWrzPuWJDIy0iZINGjQAIA022RxHaRPnz4NABgxYkSJj5uamoq8vDykpaWhWbNmd10fEdln4MCBGDNmDH777Ter0yh3WrVqFQwGAzZu3Gg1gnPJkiU266rVanTt2hVdu3bFhx9+iHfffRevv/46tm3bhm7dusnrDRgwAD169MDIkSPh5eWF+fPny7dFRERgy5Yt6NixY6lhIDQ0FIB0fAkPD5eXJyYm2oxWKknnzp0RFBSE5cuXo1OnTvj555/x+uuvW60TERGBQ4cOoWvXrvf8JarwWFjUqVOnrDr6UulcIsAA0jeMp556CtevX7frG0bnzp0xb948BAUFQavVYsmSJVbDsQu/YWzbtg0//fQTNmzYgOXLlyM2NhabNm2CRqOR13344Yfx7bff4v3338eXX35pFXAKv2EsW7as2HoCAgIq5PUHBgbi4MGD2LhxI9avX4/169djyZIlePzxx/HFF18AKLlV485vTPei8BvW9OnTrVqiivL09ERycnKFPScRlc7T0xPz589HfHw8+vfvX+J6Go0GKpXK6pgQHx+P1atXW62XnJwMPz8/q2WF7/c7hxsDwOOPP460tDQ8//zzMJlMmDZtGgCp9XzevHl466238O6771rdp6CgABkZGfDx8UG3bt2g1Woxe/Zs9OjRQz6WzZw5095NALVajcceewyfffYZ2rZti4KCApuRRYMHD8a6deuwePFiPP3001a3ZWdnw2KxwMPDw67nW716Na5cuSL3g/njjz/w+++/292qTi4UYPgNA9DpdOjfvz/69+8Pi8WCsWPHYuHChXjjjTcQGRkpt+SkpKRYBbzC1p87nTlzBkIIq+Bz6tQpACjxW0RERAQA6XRT0W10p4CAAJhMpjLnReCpJKKKUVqraKG+ffviww8/RK9evTBs2DAkJCRg7ty5iIyMxOHDh+X1pk6dih07dqBv374IDQ1FQkIC5s2bh9q1a6NTp07FPvZzzz2HtLQ0vP766/D29sZrr72GmJgYjBkzBu+99x4OHjyIHj16QKvV4vTp01ixYgVmzZqFxx57DAEBAXjppZfw3nvvoV+/fujTpw8OHDiA9evXw9/f3+5tEBcXh9mzZ2PSpElo3ry5zYCA4cOH47vvvsMzzzyDbdu2oWPHjjCbzThx4gS+++47bNy4sdhT3sWJjIxEp06d8OyzzyI3NxczZ85EjRo1MHHiRLvrdXkK98GpNHd2UhNCiM8//1xMnjzZqpPunZ3U/u///k+4u7uLzMxMedn58+eFu7u7VQfXmzdv2jznTz/9JACItWvXCiFsh1HPnj1bABATJ06U77N9+3YBwGr4daH8/Hy5M11CQoLQarV33Yk3KSnJZtncuXMFAHHkyBEhhBBr164VAMSPP/4or5ORkSHq1q1brk68rVq1kpfd2YnXbDaLiIgIUb9+fZGenm5TU0JCgvx/ezrxrl+/XgAQP/zwQ6mvn4huK+74WJziOvF++umnon79+kKv14tGjRqJJUuW2AwA2Lp1q3j44YdFcHCw0Ol0Ijg4WAwdOlScOnVKXqe4aSaEEGLixIkCgJgzZ468bNGiRSIqKkoYjUbh5eUlmjdvLiZOnCiuXr0qr2M2m8WUKVNEUFDQXU0zIYR0XKlTp44AIN5+++1i18nLyxPTpk0TTZs2FXq9Xvj6+oqoqCgxZcoUkZqaKq8HQIwbN87m/oWdeKdPny5mzJgh6tSpI/R6vYiOjhaHDh2yq06SqIQo0vOqGvn8888xatQo7N27t9REHBYWhmbNmmHt2rUApCG+Xbt2RXR0tNU3jFq1auHw4cNyR7UXX3yx2G8YKpUKR44cgbe3N7Zv344HH3wQK1askPvcFLbSvPPOO/KslM888wwWLlyI3r17l/gNA5CGCr/33nvo06eP1TeMvLw89O3bF59//nmJr3PgwIFITk5GbGwsateujQsXLmD27NkICwvDvn37oFarkZ+fj8jISGRlZWHChAnQaDT47LPPYDQasW/fPpw/f15uWQkLC4Ner0dCQgKeeeYZ1KxZE5999hmOHj2KdevWoWfPngAgb4Nt27ahS5cu8rLevXsjMDAQo0aNQkhICK5cuYJt27bBZDJhzZo1AIArV67g/vvvR1paGp5++mk0btwY165dw4oVK7Br1y74+Pjg+vXrqF27Ntq0aYNnnnkGer0esbGxJfYpIiKiakLhAFVp+A3D2sqVK0WPHj1EYGCg0Ol0om7dumLMmDHi2rVrVuvt27dPPPDAA/I6H374YZkT2bVo0ULeVne+1pImsjtw4IB45JFHRI0aNYRerxehoaFi8ODBYuvWrVbrXbhwQTz++OMiICBA6PV6ER4eLsaNGydyc3PldRYvXizCw8OFRqPhkGoiIhdRbVtgyDFs3boV3bp1w86dO0s8901ERFReLjEPDCnn2rVrAFCujnRERERlYQsMVYrMzEwsW7YMs2bNQlpaGi5cuMALOhIRUYXhJwpVisTERDz//PMwGo1YtWoVwwsREVUotsAQERGR0+HXYiIiInI6DDBERETkdBhgiIiIyOkwwBAREZHTYYAhIiIip8MAQ0RERE6HAYaIiIicDgMMEREROR0GGCIiInI6/w+OskYpFPoMfAAAAABJRU5ErkJggg==\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "# Visualize the overlap of the results using Venn diagrams\n",
        "import matplotlib.pyplot as plt\n",
        "from matplotlib_venn import venn2\n",
        "\n",
        "# accurate predictions from masked subject experiments:\n",
        "subjects_correct = filtered_results['full_sentence'].where(filtered_results['subject_is_accurate'] == True).tolist()\n",
        "verbs_correct = filtered_results['full_sentence'].where(filtered_results['verb_is_accurate'] == True).tolist()\n",
        "\n",
        "venn2([set(subjects_correct), set(verbs_correct)], set_colors=['blue', 'cyan'], set_labels=['Masked subject', 'Masked verb'])\n",
        "plt.title('Overlap between accurate results of masked subjects and verbs:')\n",
        "plt.show()\n",
        "\n",
        "subjects_incorrect = filtered_results.index[filtered_results['subject_is_accurate'] == False].tolist()\n",
        "verbs_incorrect = filtered_results.index[filtered_results['verb_is_accurate'] == False].tolist()\n",
        "venn2([set(subjects_incorrect), set(verbs_incorrect)], set_colors=['red', 'orange'], set_labels=['Masked subject', 'Masked verb'])\n",
        "plt.title('Overlap between negative results of masked subjects and verbs:')\n",
        "plt.show()\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 467
        },
        "id": "EDcau6kLJDmT",
        "outputId": "c9ac8877-02aa-4aa0-abda-f2b19bdcc2f8"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.legend.Legend at 0x7ff50d623520>"
            ]
          },
          "metadata": {},
          "execution_count": 17
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "import seaborn as sns\n",
        "\n",
        "sns.pointplot(filtered_results, x='attractors', y='verb_is_accurate', errorbar=None, color='cyan', label='Verb')\n",
        "ax = sns.pointplot(filtered_results, x='attractors', y='subject_is_accurate', errorbar=None, label='Subject')\n",
        "ax.set(ylabel='Accuracy', xlabel='Attractors')\n",
        "ax.legend()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3rakAAVJKq0A"
      },
      "source": [
        "## <a name=\"nonce-dataset-results\"></a>Results of the nonce dataset"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "%%capture\n",
        "# Reload the results so we don't need to run the model inference every time\n",
        "! [ ! -f number_agreement_nonce_with_results.csv ] && wget https://raw.githubusercontent.com/tsido/LDA-T307-project/main/data/number_agreement_nonce_with_results.csv\n",
        "\n",
        "import pandas as pd\n",
        "nonce_df = pd.read_csv('number_agreement_nonce_with_results.csv')"
      ],
      "metadata": {
        "id": "WIOKNo_JIYrO"
      },
      "execution_count": 22,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": 23,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "e5TYEh8o5M14",
        "outputId": "15bdedb3-150e-4cff-e742-a5c39889677e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Shared logits for subjects 80\n",
            "Shared logits for verbs 5\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "verb_is_accurate       0.817518\n",
              "subject_is_accurate    0.795620\n",
              "dtype: float64"
            ]
          },
          "metadata": {},
          "execution_count": 23
        }
      ],
      "source": [
        "unfiltered_nonce_results = nonce_df\n",
        "print('Shared logits for subjects', len(unfiltered_nonce_results.query('subject_true_label_p == subject_false_label_p')))\n",
        "print('Shared logits for verbs', len(unfiltered_nonce_results.query('verb_true_label_p == verb_false_label_p')))\n",
        "\n",
        "filtered_nonce_results = nonce_df[(nonce_df['subject_true_label_p'] != nonce_df['subject_false_label_p']) & (nonce_df['verb_true_label_p'] != nonce_df['verb_false_label_p'])]\n",
        "results = filtered_nonce_results[['verb_is_accurate', 'subject_is_accurate']].mean()\n",
        "results\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 24,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1480
        },
        "id": "fUDDmNsqD0Gt",
        "outputId": "0f983355-0229-4907-eb7d-38127b273f49"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Unnamed: 0.1</th>\n",
              "      <th>Unnamed: 0</th>\n",
              "      <th>full_sentence</th>\n",
              "      <th>verb_masked_sentence</th>\n",
              "      <th>verb_true_label</th>\n",
              "      <th>verb_false_label</th>\n",
              "      <th>verb_true_label_p</th>\n",
              "      <th>verb_false_label_p</th>\n",
              "      <th>verb_is_accurate</th>\n",
              "      <th>subject</th>\n",
              "      <th>subject_masked_sentence</th>\n",
              "      <th>subject_true_label</th>\n",
              "      <th>subject_false_label</th>\n",
              "      <th>subject_true_label_p</th>\n",
              "      <th>subject_false_label_p</th>\n",
              "      <th>subject_is_accurate</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>54</th>\n",
              "      <td>55</td>\n",
              "      <td>126</td>\n",
              "      <td>the electoral parties we tolerate in our label...</td>\n",
              "      <td>the electoral parties we tolerate in our label...</td>\n",
              "      <td>assume</td>\n",
              "      <td>assumes</td>\n",
              "      <td>-5.141294</td>\n",
              "      <td>-4.430453</td>\n",
              "      <td>False</td>\n",
              "      <td>parties</td>\n",
              "      <td>the electoral [MASK] we tolerate in our label ...</td>\n",
              "      <td>parties</td>\n",
              "      <td>party</td>\n",
              "      <td>5.001207</td>\n",
              "      <td>5.211483</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>56</th>\n",
              "      <td>57</td>\n",
              "      <td>130</td>\n",
              "      <td>Because of the spreadsheet that his hostile to...</td>\n",
              "      <td>Because of the spreadsheet that his hostile to...</td>\n",
              "      <td>ensures</td>\n",
              "      <td>ensure</td>\n",
              "      <td>-2.902240</td>\n",
              "      <td>-0.632169</td>\n",
              "      <td>False</td>\n",
              "      <td>hostile</td>\n",
              "      <td>Because of the spreadsheet that his [MASK] tom...</td>\n",
              "      <td>hostile</td>\n",
              "      <td>hostiles</td>\n",
              "      <td>-2.589787</td>\n",
              "      <td>-0.680003</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>66</th>\n",
              "      <td>67</td>\n",
              "      <td>150</td>\n",
              "      <td>the electrostatic fisheries we develop in our ...</td>\n",
              "      <td>the electrostatic fisheries we develop in our ...</td>\n",
              "      <td>exist</td>\n",
              "      <td>exists</td>\n",
              "      <td>-1.461376</td>\n",
              "      <td>-0.624074</td>\n",
              "      <td>False</td>\n",
              "      <td>fisheries</td>\n",
              "      <td>the electrostatic [MASK] we develop in our mon...</td>\n",
              "      <td>fisheries</td>\n",
              "      <td>fishery</td>\n",
              "      <td>-0.124862</td>\n",
              "      <td>0.933351</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>126</th>\n",
              "      <td>129</td>\n",
              "      <td>282</td>\n",
              "      <td>The historical voting movements wash there ope...</td>\n",
              "      <td>The historical voting movements wash there [MA...</td>\n",
              "      <td>operates</td>\n",
              "      <td>operate</td>\n",
              "      <td>-4.160288</td>\n",
              "      <td>-3.440489</td>\n",
              "      <td>False</td>\n",
              "      <td>movements</td>\n",
              "      <td>The historical voting [MASK] wash there operat...</td>\n",
              "      <td>movements</td>\n",
              "      <td>movement</td>\n",
              "      <td>-0.854700</td>\n",
              "      <td>1.717710</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>172</th>\n",
              "      <td>178</td>\n",
              "      <td>388</td>\n",
              "      <td>The interested popularity to reconsider for la...</td>\n",
              "      <td>The interested popularity to reconsider for la...</td>\n",
              "      <td>ensures</td>\n",
              "      <td>ensure</td>\n",
              "      <td>-2.401338</td>\n",
              "      <td>-1.728330</td>\n",
              "      <td>False</td>\n",
              "      <td>ensures</td>\n",
              "      <td>The interested popularity to reconsider for la...</td>\n",
              "      <td>ensures</td>\n",
              "      <td>ensure</td>\n",
              "      <td>-2.401338</td>\n",
              "      <td>-1.728330</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>181</th>\n",
              "      <td>187</td>\n",
              "      <td>408</td>\n",
              "      <td>the academic grandparents we proceed in our ch...</td>\n",
              "      <td>the academic grandparents we proceed in our ch...</td>\n",
              "      <td>enjoy</td>\n",
              "      <td>enjoys</td>\n",
              "      <td>-0.036061</td>\n",
              "      <td>0.482687</td>\n",
              "      <td>False</td>\n",
              "      <td>grandparents</td>\n",
              "      <td>the academic [MASK] we proceed in our channel ...</td>\n",
              "      <td>grandparents</td>\n",
              "      <td>grandparent</td>\n",
              "      <td>-5.555443</td>\n",
              "      <td>1.407114</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>282</th>\n",
              "      <td>291</td>\n",
              "      <td>642</td>\n",
              "      <td>The joint lot cases recommend there adds raid ...</td>\n",
              "      <td>The joint lot cases recommend there [MASK] rai...</td>\n",
              "      <td>adds</td>\n",
              "      <td>add</td>\n",
              "      <td>0.311989</td>\n",
              "      <td>2.909750</td>\n",
              "      <td>False</td>\n",
              "      <td>cases</td>\n",
              "      <td>The joint lot [MASK] recommend there adds raid...</td>\n",
              "      <td>cases</td>\n",
              "      <td>case</td>\n",
              "      <td>0.075195</td>\n",
              "      <td>0.837249</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>286</th>\n",
              "      <td>295</td>\n",
              "      <td>650</td>\n",
              "      <td>My yoga with answering the throat technically ...</td>\n",
              "      <td>My yoga with answering the throat technically ...</td>\n",
              "      <td>observes</td>\n",
              "      <td>observe</td>\n",
              "      <td>-3.396935</td>\n",
              "      <td>-0.763799</td>\n",
              "      <td>False</td>\n",
              "      <td>yoga</td>\n",
              "      <td>My [MASK] with answering the throat technicall...</td>\n",
              "      <td>yoga</td>\n",
              "      <td>yogas</td>\n",
              "      <td>-1.642658</td>\n",
              "      <td>-1.042765</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>356</th>\n",
              "      <td>366</td>\n",
              "      <td>814</td>\n",
              "      <td>The obsessed coding I built actually from them...</td>\n",
              "      <td>The obsessed coding I built actually from them...</td>\n",
              "      <td>extends</td>\n",
              "      <td>extend</td>\n",
              "      <td>-2.540220</td>\n",
              "      <td>-2.483209</td>\n",
              "      <td>False</td>\n",
              "      <td>coding</td>\n",
              "      <td>The obsessed [MASK] I built actually from them...</td>\n",
              "      <td>coding</td>\n",
              "      <td>codings</td>\n",
              "      <td>-2.048849</td>\n",
              "      <td>1.359644</td>\n",
              "      <td>False</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "application/vnd.google.colaboratory.module+javascript": "\n      import \"https://ssl.gstatic.com/colaboratory/data_table/99dac6621f6ae8c4/data_table.js\";\n\n      window.createDataTable({\n        data: [[{\n            'v': 55,\n            'f': \"55\",\n        },\n{\n            'v': 126,\n            'f': \"126\",\n        },\n\"the electoral parties we tolerate in our label assume ; we manage percentage tutorial for easter , climate chili and warning - rein for Christianity . <eos>\",\n\"the electoral parties we tolerate in our label [MASK] ; we manage percentage tutorial for easter , climate chili and warning - rein for Christianity . <eos>\",\n\"assume\",\n\"assumes\",\n{\n            'v': -5.141293525695801,\n            'f': \"-5.141293525695801\",\n        },\n{\n            'v': -4.430453300476074,\n            'f': \"-4.430453300476074\",\n        },\nfalse,\n\"parties\",\n\"the electoral [MASK] we tolerate in our label assume ; we manage percentage tutorial for easter , climate chili and warning - rein for Christianity . <eos>\",\n\"parties\",\n\"party\",\n{\n            'v': 5.001206874847412,\n            'f': \"5.001206874847412\",\n        },\n{\n            'v': 5.211483478546143,\n            'f': \"5.211483478546143\",\n        },\nfalse],\n [{\n            'v': 57,\n            'f': \"57\",\n        },\n{\n            'v': 130,\n            'f': \"130\",\n        },\n\"Because of the spreadsheet that his hostile tomorrow with Janet Alaska had tended him , he waited cm necessary as : \\\" You can foresee a hour and a thermometer , eighteen requirement you ca n't come ensures why you 're visual effectively \\\" and belonged courses to the \\\" Baudelaire Brazil \\\" team preventing that \\\" potentially he refused Few policies to settle just western as a foster \\\" . <eos>\",\n\"Because of the spreadsheet that his hostile tomorrow with Janet Alaska had tended him , he waited cm necessary as : \\\" You can foresee a hour and a thermometer , eighteen requirement you ca n't come [MASK] why you 're visual effectively \\\" and belonged courses to the \\\" Baudelaire Brazil \\\" team preventing that \\\" potentially he refused Few policies to settle just western as a foster \\\" . <eos>\",\n\"ensures\",\n\"ensure\",\n{\n            'v': -2.90224027633667,\n            'f': \"-2.90224027633667\",\n        },\n{\n            'v': -0.6321690678596497,\n            'f': \"-0.6321690678596497\",\n        },\nfalse,\n\"hostile\",\n\"Because of the spreadsheet that his [MASK] tomorrow with Janet Alaska had tended him , he waited cm necessary as : \\\" You can foresee a hour and a thermometer , eighteen requirement you ca n't come ensures why you 're visual effectively \\\" and belonged courses to the \\\" Baudelaire Brazil \\\" team preventing that \\\" potentially he refused Few policies to settle just western as a foster \\\" . <eos>\",\n\"hostile\",\n\"hostiles\",\n{\n            'v': -2.589787006378174,\n            'f': \"-2.589787006378174\",\n        },\n{\n            'v': -0.6800034046173096,\n            'f': \"-0.6800034046173096\",\n        },\nfalse],\n [{\n            'v': 67,\n            'f': \"67\",\n        },\n{\n            'v': 150,\n            'f': \"150\",\n        },\n\"the electrostatic fisheries we develop in our monkey exist ; we listen perspective wall for easter , planet testimony and supervisor - budget for Wii . <eos>\",\n\"the electrostatic fisheries we develop in our monkey [MASK] ; we listen perspective wall for easter , planet testimony and supervisor - budget for Wii . <eos>\",\n\"exist\",\n\"exists\",\n{\n            'v': -1.4613759517669678,\n            'f': \"-1.4613759517669678\",\n        },\n{\n            'v': -0.6240735054016113,\n            'f': \"-0.6240735054016113\",\n        },\nfalse,\n\"fisheries\",\n\"the electrostatic [MASK] we develop in our monkey exist ; we listen perspective wall for easter , planet testimony and supervisor - budget for Wii . <eos>\",\n\"fisheries\",\n\"fishery\",\n{\n            'v': -0.1248621121048927,\n            'f': \"-0.1248621121048927\",\n        },\n{\n            'v': 0.9333506226539612,\n            'f': \"0.9333506226539612\",\n        },\nfalse],\n [{\n            'v': 129,\n            'f': \"129\",\n        },\n{\n            'v': 282,\n            'f': \"282\",\n        },\n\"The historical voting movements wash there operates breathe to the passport . <eos>\",\n\"The historical voting movements wash there [MASK] breathe to the passport . <eos>\",\n\"operates\",\n\"operate\",\n{\n            'v': -4.160288333892822,\n            'f': \"-4.160288333892822\",\n        },\n{\n            'v': -3.440488815307617,\n            'f': \"-3.440488815307617\",\n        },\nfalse,\n\"movements\",\n\"The historical voting [MASK] wash there operates breathe to the passport . <eos>\",\n\"movements\",\n\"movement\",\n{\n            'v': -0.8547003269195557,\n            'f': \"-0.8547003269195557\",\n        },\n{\n            'v': 1.71770977973938,\n            'f': \"1.71770977973938\",\n        },\nfalse],\n [{\n            'v': 178,\n            'f': \"178\",\n        },\n{\n            'v': 388,\n            'f': \"388\",\n        },\n\"The interested popularity to reconsider for lax ensures to retrieve six , or your professor difference criticism competition , and adjust . <eos>\",\n\"The interested popularity to reconsider for lax [MASK] to retrieve six , or your professor difference criticism competition , and adjust . <eos>\",\n\"ensures\",\n\"ensure\",\n{\n            'v': -2.4013381004333496,\n            'f': \"-2.4013381004333496\",\n        },\n{\n            'v': -1.728330135345459,\n            'f': \"-1.728330135345459\",\n        },\nfalse,\n\"ensures\",\n\"The interested popularity to reconsider for lax [MASK] to retrieve six , or your professor difference criticism competition , and adjust . <eos>\",\n\"ensures\",\n\"ensure\",\n{\n            'v': -2.4013381004333496,\n            'f': \"-2.4013381004333496\",\n        },\n{\n            'v': -1.728330135345459,\n            'f': \"-1.728330135345459\",\n        },\nfalse],\n [{\n            'v': 187,\n            'f': \"187\",\n        },\n{\n            'v': 408,\n            'f': \"408\",\n        },\n\"the academic grandparents we proceed in our channel enjoy ; we pose market none for easter , flipping reputation and scam - tonight for Chechnya . <eos>\",\n\"the academic grandparents we proceed in our channel [MASK] ; we pose market none for easter , flipping reputation and scam - tonight for Chechnya . <eos>\",\n\"enjoy\",\n\"enjoys\",\n{\n            'v': -0.0360608398914337,\n            'f': \"-0.0360608398914337\",\n        },\n{\n            'v': 0.4826874136924743,\n            'f': \"0.4826874136924743\",\n        },\nfalse,\n\"grandparents\",\n\"the academic [MASK] we proceed in our channel enjoy ; we pose market none for easter , flipping reputation and scam - tonight for Chechnya . <eos>\",\n\"grandparents\",\n\"grandparent\",\n{\n            'v': -5.555442810058594,\n            'f': \"-5.555442810058594\",\n        },\n{\n            'v': 1.4071143865585327,\n            'f': \"1.4071143865585327\",\n        },\nfalse],\n [{\n            'v': 291,\n            'f': \"291\",\n        },\n{\n            'v': 642,\n            'f': \"642\",\n        },\n\"The joint lot cases recommend there adds raid to the craft . <eos>\",\n\"The joint lot cases recommend there [MASK] raid to the craft . <eos>\",\n\"adds\",\n\"add\",\n{\n            'v': 0.3119890689849853,\n            'f': \"0.3119890689849853\",\n        },\n{\n            'v': 2.909749746322632,\n            'f': \"2.909749746322632\",\n        },\nfalse,\n\"cases\",\n\"The joint lot [MASK] recommend there adds raid to the craft . <eos>\",\n\"cases\",\n\"case\",\n{\n            'v': 0.0751950964331626,\n            'f': \"0.0751950964331626\",\n        },\n{\n            'v': 0.8372493386268616,\n            'f': \"0.8372493386268616\",\n        },\nfalse],\n [{\n            'v': 295,\n            'f': \"295\",\n        },\n{\n            'v': 650,\n            'f': \"650\",\n        },\n\"My yoga with answering the throat technically observes , could it be understood that properly everyone belonged the parent would or could establish at any swing , since we correctly warned Courtney and Californians and if they killed they might have saved a cruel violence ?? <eos>\",\n\"My yoga with answering the throat technically [MASK] , could it be understood that properly everyone belonged the parent would or could establish at any swing , since we correctly warned Courtney and Californians and if they killed they might have saved a cruel violence ?? <eos>\",\n\"observes\",\n\"observe\",\n{\n            'v': -3.396935224533081,\n            'f': \"-3.396935224533081\",\n        },\n{\n            'v': -0.763798713684082,\n            'f': \"-0.763798713684082\",\n        },\nfalse,\n\"yoga\",\n\"My [MASK] with answering the throat technically observes , could it be understood that properly everyone belonged the parent would or could establish at any swing , since we correctly warned Courtney and Californians and if they killed they might have saved a cruel violence ?? <eos>\",\n\"yoga\",\n\"yogas\",\n{\n            'v': -1.642658233642578,\n            'f': \"-1.642658233642578\",\n        },\n{\n            'v': -1.0427650213241575,\n            'f': \"-1.0427650213241575\",\n        },\nfalse],\n [{\n            'v': 366,\n            'f': \"366\",\n        },\n{\n            'v': 814,\n            'f': \"814\",\n        },\n\"The obsessed coding I built actually from them in the first chair extends because my distinguishing security who owned with them was returning out on his unknown . <eos>\",\n\"The obsessed coding I built actually from them in the first chair [MASK] because my distinguishing security who owned with them was returning out on his unknown . <eos>\",\n\"extends\",\n\"extend\",\n{\n            'v': -2.5402204990386963,\n            'f': \"-2.5402204990386963\",\n        },\n{\n            'v': -2.4832093715667725,\n            'f': \"-2.4832093715667725\",\n        },\nfalse,\n\"coding\",\n\"The obsessed [MASK] I built actually from them in the first chair extends because my distinguishing security who owned with them was returning out on his unknown . <eos>\",\n\"coding\",\n\"codings\",\n{\n            'v': -2.04884934425354,\n            'f': \"-2.04884934425354\",\n        },\n{\n            'v': 1.3596444129943848,\n            'f': \"1.3596444129943848\",\n        },\nfalse]],\n        columns: [[\"number\", \"Unnamed: 0.1\"], [\"number\", \"Unnamed: 0\"], [\"string\", \"full_sentence\"], [\"string\", \"verb_masked_sentence\"], [\"string\", \"verb_true_label\"], [\"string\", \"verb_false_label\"], [\"number\", \"verb_true_label_p\"], [\"number\", \"verb_false_label_p\"], [\"string\", \"verb_is_accurate\"], [\"string\", \"subject\"], [\"string\", \"subject_masked_sentence\"], [\"string\", \"subject_true_label\"], [\"string\", \"subject_false_label\"], [\"number\", \"subject_true_label_p\"], [\"number\", \"subject_false_label_p\"], [\"string\", \"subject_is_accurate\"]],\n        columnOptions: [],\n        rowsPerPage: 5,\n        helpUrl: \"https://colab.research.google.com/notebooks/data_table.ipynb\",\n        suppressOutputScrolling: true,\n        minimumWidth: undefined,\n      });\n    ",
            "text/plain": [
              "<google.colab.data_table.DataTable object>"
            ]
          },
          "metadata": {},
          "execution_count": 24
        }
      ],
      "source": [
        "data = filtered_nonce_results[(filtered_nonce_results['subject_is_accurate'] == False) & (filtered_nonce_results['verb_is_accurate'] == False)]\n",
        "data_table.DataTable(data, include_index=False, num_rows_per_page=5)\n"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Visualize the overlap of the results using Venn diagrams\n",
        "import matplotlib.pyplot as plt\n",
        "from matplotlib_venn import venn2\n",
        "\n",
        "# accurate predictions from masked subject experiments for nonce senntences:\n",
        "subjects_correct = filtered_nonce_results['full_sentence'].where(filtered_nonce_results['subject_is_accurate'] == True).tolist()\n",
        "verbs_correct = filtered_nonce_results['full_sentence'].where(filtered_nonce_results['verb_is_accurate'] == True).tolist()\n",
        "\n",
        "venn2([set(subjects_correct), set(verbs_correct)], set_colors=['blue', 'cyan'], set_labels=['Masked subject', 'Masked verb'])\n",
        "plt.title('Overlap between accurate results of masked subjects and verbs for nonce dataset:')\n",
        "plt.show()\n",
        "\n",
        "subjects_incorrect = filtered_nonce_results.index[filtered_nonce_results['subject_is_accurate'] == False].tolist()\n",
        "verbs_incorrect = filtered_nonce_results.index[filtered_nonce_results['verb_is_accurate'] == False].tolist()\n",
        "venn2([set(subjects_incorrect), set(verbs_incorrect)], set_colors=['red', 'orange'], set_labels=['Masked subject', 'Masked verb'])\n",
        "plt.title('Overlap between negative results of masked subjects and verbs for nonce dataset:')\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 795
        },
        "id": "6OW-7SG1D3Bp",
        "outputId": "4b32a3e0-6317-4841-ab05-545220693f7f"
      },
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "s_Kx7rIJB7IA"
      },
      "source": [
        "## Discussion of the results\n",
        "\n",
        "* The errors for verbs and subjects are mainly made in different sentences. Of total of 2259 errors only 93 are made in both subject and verb case in the large dataset and of 97 total errors only 9 are shared in the nonce dataset.\n",
        "* The correct masked verb is predicted with a slighty higher accuracy than subject in both datasets. \n",
        "* The number of attractors lowers the accuracy in the case of subjects\n",
        "* The probabilities for many of the candidate subjects and verbs are equal for the true and false labels. This is because words are not found as individual tokens and are labeled as `[UNK]` tokens instead. The vast majority of these cases are in the masked subject experiments.  \n",
        "  * In the large dataset:\n",
        "    * There are 103 (5.4%) cases of these for verbs\n",
        "    * And 1817 (94.6%) cases for subjects\n",
        "  * In the nonce dataset:\n",
        "    * 5 (5.9%) cases for verbs\n",
        "    * 80 (94.1%) for subjects\n",
        "  * These uknown tokens likely skew the results somewhat\n",
        "* Another issue with the masked subject test is that the subjects are parsed programmatically and there hasn't been a\n",
        "separate step to validate that the parsing is done accurately. \n",
        "* The results for masked verbs are few percent lower than in the original experiment. A different model was used, `bert-base-cased` vs `bert-base-uncased`, that could explain it.\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PUJ_JlYOgPsB"
      },
      "source": [
        "## Conclusions \n",
        "\n",
        "The results of the experiment are inline with the original experiments by Goldberg (2019). The accuracy for the masked subject experiment are slightly lower than the results for the masked verbs and the errors seem to be located in different sentences. Further drilling down on the error patterns could give insight on why this is. There is an issue especially with the masked subject test, as many of the subjects are tokenized with an unknown token id, which can skew the results. The experiment could be expanded to other languages. For example for Finnish the [FinBERT](https://www.kielipankki.fi/tools/finbert/) could be used as the model and the [Finnish TreeBank (FTB)](https://www.kielipankki.fi/corpora/finntreebank/) could be used for generating the datasets.  \n"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "provenance": []
    },
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.10.8"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}