wsd.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [],
   "source": [
    "from html.parser import HTMLParser\n",
    "import os\n",
    "import math\n",
    "import string\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "pd.set_option('display.max_colwidth', 200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import nltk\n",
    "# nltk.download()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from nltk.tokenize import word_tokenize\n",
    "from nltk.stem import PorterStemmer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "class BNCParser(HTMLParser):\n",
    "    def __init__(self):\n",
    "        HTMLParser.__init__(self)\n",
    "        self.parsed_instances = list()\n",
    "        self.recording = False\n",
    "        self.this_instance = dict()\n",
    "        self.this_data = list()\n",
    "        self.ps = PorterStemmer()\n",
    "        self.stopwords = nltk.corpus.stopwords.words('english')\n",
    "        \n",
    "    def handle_starttag(self, tag, attrs):\n",
    "        if tag == 'instance':\n",
    "            self.this_instance['id'] = attrs[0][1]\n",
    "        if tag == 'context':\n",
    "            self.recording =  True\n",
    "    \n",
    "    def handle_startendtag(self, tag, attrs):\n",
    "        if tag == 'answer':\n",
    "            self.this_instance['sense'] = attrs[1][1].split('%')[1]\n",
    "            \n",
    "    def handle_data(self, data):\n",
    "        if self.recording:\n",
    "            self.this_data.append(data)\n",
    "            \n",
    "    def handle_endtag(self, tag):\n",
    "        if tag == 'context':\n",
    "            self.recording = False\n",
    "        if tag == 'instance':\n",
    "            self.this_data = ' '.join(self.this_data)\n",
    "            self.this_data = self.preprocess_text(self.this_data)\n",
    "            self.this_instance['data'] = self.this_data\n",
    "            self.parsed_instances.append(self.this_instance)\n",
    "\n",
    "            self.this_instance = dict()\n",
    "            self.this_data = list()\n",
    "            \n",
    "    def preprocess_text(self, txt):\n",
    "        # Removes punctuation, tokenizes and stems all tokens\n",
    "        txt_without_punc = txt.lower().translate(str.maketrans('', '', string.punctuation))\n",
    "        tokenized_txt = [w for w in word_tokenize(txt_without_punc) if w not in self.stopwords]\n",
    "        stemmed_tokens = [self.ps.stem(w) for w in tokenized_txt]\n",
    "        stemmed_tokens = [w for w in stemmed_tokens if not w.isdigit()]\n",
    "        return stemmed_tokens"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_f_given_s_df(train, vocab_size, disambiguate_word):\n",
    "    f_given_s = (\n",
    "        train\n",
    "        .assign(sense_count = lambda x: x.groupby('sense')['sense'].transform('count'))\n",
    "        .explode('data')\n",
    "        .groupby(['data', 'sense', 'sense_count'])\n",
    "        .agg({'data': 'count'})\n",
    "        .rename(columns={'data': 'count'})\n",
    "        .reset_index()\n",
    "        .sort_values(by='count', ascending=False)\n",
    "        .loc[lambda x: x['data'] != disambiguate_word]\n",
    "        .assign(\n",
    "            smooth_count = lambda x: x['count'] + 1,\n",
    "            smooth_sense_count = lambda x: x['sense_count'] + vocab_size\n",
    "        )\n",
    "        .assign(prob = lambda x: np.log(x['smooth_count']/x['smooth_sense_count']))\n",
    "        .set_index(['data', 'sense'])\n",
    "        ['prob']\n",
    "    )\n",
    "    return f_given_s"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_f_given_s_prob(f, s, f_given_s, sense_count_dict, vocab_size):\n",
    "    try:\n",
    "        return f_given_s[(f, s)]\n",
    "    except KeyError:\n",
    "        return np.log(1.0/(sense_count_dict[s] + vocab_size))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_s_prob(s, sense_count_dict):\n",
    "    return np.log(sense_count_dict[s]/sum(sense_count_dict.values()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_vocab_size(df):\n",
    "    list_of_list = df['data'].values\n",
    "    return len(set([word for lst in list_of_list for word in lst]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict_sense(word_list, f_given_s, sense_count_dict, vocab_size):\n",
    "    val = dict()\n",
    "    for s in sense_count_dict.keys():\n",
    "        current_val = 0\n",
    "        for word in word_list:\n",
    "            current_val += get_f_given_s_prob(word, s, f_given_s, \n",
    "                                              sense_count_dict, vocab_size)\n",
    "        val[s] = current_val + get_s_prob(s, sense_count_dict)\n",
    "    return max(val, key=val.get), val"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 175,
   "metadata": {},
   "outputs": [],
   "source": [
    "def split_data(df, folds):\n",
    "    n = len(df)\n",
    "    elem_count = [math.ceil(n/folds)] * (folds - 1) + [n - (folds - 1) * math.ceil(n/folds)]\n",
    "    \n",
    "    chunks = []\n",
    "    for count in elem_count:\n",
    "        chunks.append(df[:count])\n",
    "        df = df.iloc[count:, :]\n",
    "    return chunks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 176,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_train_test(split_data, test_index):\n",
    "    test = split_data[test_index]\n",
    "    train_data = [split_data[i] for i in range(len(split_data)) \n",
    "                  if i != test_index]\n",
    "    \n",
    "    train = pd.DataFrame()\n",
    "    for data in train_data:\n",
    "        train = train.append(data)\n",
    "    return train, test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 177,
   "metadata": {},
   "outputs": [],
   "source": [
    "def accuracy_score(s1, s2):\n",
    "    correct = np.sum(s1 == s2)\n",
    "    total = len(s1)\n",
    "    acc = round((correct/total) * 100, 2)\n",
    "    return correct, total, acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 178,
   "metadata": {},
   "outputs": [],
   "source": [
    "def write_test_file(test, disambiguate_word, file):\n",
    "    test['output'] = disambiguate_word + '%' + test['prediction']\n",
    "    test = test[['id', 'output']]\n",
    "    test.to_csv(file, index=False, sep=' ', header=False, mode='a')\n",
    "    return"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "-------------------"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "metadata": {},
   "outputs": [],
   "source": [
    "filename = 'plant-1.wsd'\n",
    "folds = 5\n",
    "\n",
    "disambiguate_word = filename.split('.')[0].split('-')[0]\n",
    "output_filename = '{}.wsd.out'.format(disambiguate_word)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 184,
   "metadata": {},
   "outputs": [],
   "source": [
    "parser = BNCParser()\n",
    "with open(filename, 'r') as f:\n",
    "    parser.feed(f.read())\n",
    "df = pd.DataFrame(parser.parsed_instances)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0 86.84\n",
      "1 92.11\n",
      "2 84.21\n",
      "3 92.11\n",
      "4 44.44\n",
      "Average: 80.32\n"
     ]
    }
   ],
   "source": [
    "try:\n",
    "    os.remove(output_filename)\n",
    "except FileNotFoundError:\n",
    "    pass\n",
    "\n",
    "correct = 0\n",
    "total = 0\n",
    "outdf = pd.DataFrame()\n",
    "\n",
    "with open(output_filename, 'a') as f:\n",
    "    for test_fold_index in range(folds):\n",
    "        train, test = get_train_test(split_data(df, 5), test_fold_index)\n",
    "\n",
    "        sense_count_dict = train.sense.value_counts().to_dict()\n",
    "        vocab_size = get_vocab_size(train)\n",
    "        f_given_s = get_f_given_s_df(train, vocab_size)\n",
    "\n",
    "        predict_func_with_context = lambda word_list: predict_sense(\n",
    "            word_list, \n",
    "            f_given_s, \n",
    "            sense_count_dict, \n",
    "            vocab_size\n",
    "        )\n",
    "\n",
    "        test['func_out'] = test.apply(lambda x: predict_func_with_context(x['data']), axis=1)\n",
    "        test['prediction'] = test['func_out'].str[0]\n",
    "        test['scores'] = test['func_out'].str[1]\n",
    "        \n",
    "        fold_correct, fold_total, acc = accuracy_score(test.sense, test.prediction)\n",
    "        correct += fold_correct\n",
    "        total += fold_total\n",
    "        accuracy_list.append(acc)\n",
    "        \n",
    "        outdf = outdf.append(test)\n",
    "        f.write('Fold {}\\n'.format(test_fold_index+1))\n",
    "        write_test_file(test, disambiguate_word, f)\n",
    "        print(test_fold_index, acc)\n",
    "\n",
    "    print('Average: {}'.format(round((correct/total) * 100, 2)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 187,
   "metadata": {},
   "outputs": [
    {
     "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>id</th>\n",
       "      <th>sense</th>\n",
       "      <th>data</th>\n",
       "      <th>func_out</th>\n",
       "      <th>prediction</th>\n",
       "      <th>scores</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>plant.1000004</td>\n",
       "      <td>living</td>\n",
       "      <td>[breweri, malt, ground, coars, powder, call, grist, thoroughli, mix, larg, vessel, call, mash, tun, hot, pure, water, sugar, dissolv, water, sweet, liquid, call, wort, pump, copper, flower, hop, p...</td>\n",
       "      <td>(factory, {'factory': -291.5626099781813, 'living': -291.6800433457175})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -291.5626099781813, 'living': -291.6800433457175}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>plant.1000009</td>\n",
       "      <td>living</td>\n",
       "      <td>[protech, polyesterbas, film, easili, fit, flat, glass, provid, benefit, toughen, glass, fraction, cost, fit, film, easi, glass, must, thoroughli, clean, made, dustfre, spray, water, ordinari, han...</td>\n",
       "      <td>(factory, {'factory': -219.21804918526936, 'living': -220.6999942531542})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -219.21804918526936, 'living': -220.6999942531542}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>plant.1000016</td>\n",
       "      <td>living</td>\n",
       "      <td>[mani, 60odd, speci, slipper, orchid, best, known, popular, orchid, west, could, extinct, within, year, without, tough, measur, halt, vast, illeg, trade, wild, varieti, dutch, govern, concern, nex...</td>\n",
       "      <td>(factory, {'factory': -482.706665346049, 'living': -485.31318249985753})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -482.706665346049, 'living': -485.31318249985753}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>plant.1000023</td>\n",
       "      <td>living</td>\n",
       "      <td>[didnt, know, roll, stone, start, play, river, regular, correspond, rever, minimalist, la, mont, young, introduc, dadaist, movement, fluxu, time, leonard, bernstein, scholarship, berkshir, school,...</td>\n",
       "      <td>(factory, {'factory': -451.2398071261946, 'living': -453.87001464170635})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -451.2398071261946, 'living': -453.87001464170635}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>plant.1000033</td>\n",
       "      <td>living</td>\n",
       "      <td>[ideal, farmer, would, like, grow, larva, egg, new, industri, prawn, hatcheri, grown, meet, demand, hatcheri, taiwan, alon, hatcheri, owner, face, problem, get, suppli, right, feedstuff, metamorph...</td>\n",
       "      <td>(factory, {'factory': -299.53540805717313, 'living': -300.2074255847066})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -299.53540805717313, 'living': -300.2074255847066}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>38</td>\n",
       "      <td>plant.1000041</td>\n",
       "      <td>living</td>\n",
       "      <td>[move, help, us, combat, trader, put, mani, rare, plant, risk, anoth, action, conserv, endang, plant, netherland, world, centr, plant, trade, second, one, lead, botanist, cite, secretariat, improv...</td>\n",
       "      <td>(factory, {'factory': -622.0543720341819, 'living': -623.2213303708998})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -622.0543720341819, 'living': -623.2213303708998}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>53</td>\n",
       "      <td>plant.1000057</td>\n",
       "      <td>living</td>\n",
       "      <td>[helgi, dump, flag, reveal, farm, go, back, guy, real, guid, didnt, insist, simpl, go, direct, plant, flag, regular, interv]</td>\n",
       "      <td>(factory, {'factory': -144.48932220740457, 'living': -144.69310340881904})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -144.48932220740457, 'living': -144.69310340881904}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>75</td>\n",
       "      <td>plant.1000082</td>\n",
       "      <td>living</td>\n",
       "      <td>[concern, risk, engin, plant, might, acquir, weedi, trait, escap, cultiv, effect, agricultur, environ, could, profound, recent, report, royal, commiss, environment, pollut, conclud, potato, exampl...</td>\n",
       "      <td>(factory, {'factory': -345.6128361385043, 'living': -355.15795021642583})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -345.6128361385043, 'living': -355.15795021642583}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>84</td>\n",
       "      <td>plant.1000091</td>\n",
       "      <td>living</td>\n",
       "      <td>[instanc, nercfund, project, tortois, southern, franc, local, involv, ten, year, later, tortois, conserv, project, selffund, toast, local, businessmen, embrac, local, peopl, forg, ahead, one, infl...</td>\n",
       "      <td>(factory, {'factory': -372.9256612396836, 'living': -386.5584717108951})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -372.9256612396836, 'living': -386.5584717108951}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>91</td>\n",
       "      <td>plant.1000098</td>\n",
       "      <td>living</td>\n",
       "      <td>[joseph, gill, john, pinney, harddrink, cousin, live, near, caretak, peggi, marsh, describ, dorothi, one, nicest, girl, ever, saw, arriv, month, help, run, enorm, hous, begin, wordsworth, dorothi,...</td>\n",
       "      <td>(factory, {'factory': -571.41903872444, 'living': -576.056221753028})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -571.41903872444, 'living': -576.056221753028}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>93</td>\n",
       "      <td>plant.1000100</td>\n",
       "      <td>living</td>\n",
       "      <td>[plagu, pestil, famin, prayer, grandfath, smallwe, rememb, occur, litani, book, common, prayer, plagu, pestil, famin, battl, murder, sudden, death, good, lord, deliv, us, bh, plant, sl]</td>\n",
       "      <td>(factory, {'factory': -201.55761382256713, 'living': -204.35727865852724})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -201.55761382256713, 'living': -204.35727865852724}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>97</td>\n",
       "      <td>plant.1000104</td>\n",
       "      <td>factory</td>\n",
       "      <td>[quick, snack, turn, milk, chocol, go, strict, vegan, diet, anim, product, ban, place, tea, coffe, even, fish, egg, salt, sugar, forbidden, like, leap, warm, bed, ici, sea, antarctica, idea, diet,...</td>\n",
       "      <td>(living, {'factory': -342.44769682361186, 'living': -337.67814820301095})</td>\n",
       "      <td>living</td>\n",
       "      <td>{'factory': -342.44769682361186, 'living': -337.67814820301095}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>98</td>\n",
       "      <td>plant.1000105</td>\n",
       "      <td>living</td>\n",
       "      <td>[voic, peopl, talk, left, music, play, plant, bodi, anim]</td>\n",
       "      <td>(factory, {'factory': -63.356233554716745, 'living': -63.71509982969176})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -63.356233554716745, 'living': -63.71509982969176}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>111</td>\n",
       "      <td>plant.1000119</td>\n",
       "      <td>living</td>\n",
       "      <td>[nottingham, recogn, sourc, special, contribut, academ, achiev, qualiti, peopl, work, within, interact, enquir, critic, mind, student, staff, pursuit, idea, desir, move, forward, boundari, knowled...</td>\n",
       "      <td>(factory, {'factory': -605.0678048253524, 'living': -617.5390265614095})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -605.0678048253524, 'living': -617.5390265614095}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>120</td>\n",
       "      <td>plant.1000128</td>\n",
       "      <td>living</td>\n",
       "      <td>[quick, reflex, action, caught, tompaw, short, stout, wooden, handl, steel, claw, set, end, thump, first, deal, day, yanto, work, make, sling, timber, readi, tini, bennion, crane, pick, dump, whar...</td>\n",
       "      <td>(factory, {'factory': -381.2742962105356, 'living': -391.21584447676037})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -381.2742962105356, 'living': -391.21584447676037}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>149</td>\n",
       "      <td>plant.1000157</td>\n",
       "      <td>living</td>\n",
       "      <td>[shaki, cane, chair, go, tabl, carri, home, head, mum, sat, hour, hour, calligraphi, christma, birthday, card, squar, lush, paper, clean, never, care, interest, wasnt, chore, get, knee, scrub, bru...</td>\n",
       "      <td>(factory, {'factory': -365.8661031855998, 'living': -369.73000719865075})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -365.8661031855998, 'living': -369.73000719865075}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>150</td>\n",
       "      <td>plant.1000159</td>\n",
       "      <td>living</td>\n",
       "      <td>[shall, see, posit, chang, slightli, encouragingli, sinc, challeng, given, burden, write, book, wide, circl, believ, possibl, consid, biblic, practic, principl, church, plant, enabl, us, promot, p...</td>\n",
       "      <td>(factory, {'factory': -327.46794958256646, 'living': -337.23828765893376})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -327.46794958256646, 'living': -337.23828765893376}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>153</td>\n",
       "      <td>plant.1000162</td>\n",
       "      <td>living</td>\n",
       "      <td>[mention, domesday, book, inde, evid, seem, indic, one, oldest, largest, settlement, time, domesday, time, armada, reign, elizabeth, mappleton, three, beacon, light, give, warn, invas, villag, mai...</td>\n",
       "      <td>(factory, {'factory': -325.0589929979348, 'living': -331.90590264772317})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -325.0589929979348, 'living': -331.90590264772317}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>154</td>\n",
       "      <td>plant.1000163</td>\n",
       "      <td>living</td>\n",
       "      <td>[howev, hope, take, time, read, chapter, preced, recip, menu, diet, chart, fail, deepli, impress, mani, way, highfibr, food, help, slim, come, conclus, fplan, major, slim, breakthrough, everyon, s...</td>\n",
       "      <td>(factory, {'factory': -427.39614901322733, 'living': -442.32821718825267})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -427.39614901322733, 'living': -442.32821718825267}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>155</td>\n",
       "      <td>plant.1000164</td>\n",
       "      <td>living</td>\n",
       "      <td>[light, necessari, viewer, plant, select, dimmest, suffic, interior, decor, also, fall, accept, inhabit, categori, predat, tiger, fish, requir, seclud, hide, place, use, curiowood, dens, growth, g...</td>\n",
       "      <td>(factory, {'factory': -248.26358998793822, 'living': -255.52522662956633})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -248.26358998793822, 'living': -255.52522662956633}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>156</td>\n",
       "      <td>plant.1000166</td>\n",
       "      <td>living</td>\n",
       "      <td>[right, float, food, provid, much, koi, vitamin, requir, b, bacteria, probabl, divers, group, live, organ, earth, bacteria, belong, plant, kingdom, rank, lowest, order, speci]</td>\n",
       "      <td>(factory, {'factory': -177.59520172795257, 'living': -179.8632370254072})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -177.59520172795257, 'living': -179.8632370254072}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>157</td>\n",
       "      <td>plant.1000167</td>\n",
       "      <td>living</td>\n",
       "      <td>[earli, oxygen, level, would, period, depress, order, allow, first, photosynthes, carri, evolv, nevertheless, veget, soon, start, prolifer, liverwort, moss, fern, conifer, plant, flower, appear, q...</td>\n",
       "      <td>(factory, {'factory': -427.63585758141625, 'living': -438.31243494853356})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -427.63585758141625, 'living': -438.31243494853356}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>158</td>\n",
       "      <td>plant.1000168</td>\n",
       "      <td>living</td>\n",
       "      <td>[gandhi, recogn, fact, imposs, sustain, one, bodi, without, destruct, bodi, extent, destroy, much, life, think, necessari, susten, preserv, destroy, plant, may, anim, life, food]</td>\n",
       "      <td>(factory, {'factory': -171.1696846158845, 'living': -173.833551265479})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -171.1696846158845, 'living': -173.833551265479}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>163</td>\n",
       "      <td>plant.1000173</td>\n",
       "      <td>living</td>\n",
       "      <td>[aim, reduc, swell, hair, perm, process, mean, hair, structur, protect, long, term, damag, finish, ian, style, sue, hair, use, natur, style, shape, mouss, creat, soft, natur, look, new, style, sha...</td>\n",
       "      <td>(factory, {'factory': -457.6589791996005, 'living': -475.84320887674215})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -457.6589791996005, 'living': -475.84320887674215}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>164</td>\n",
       "      <td>plant.1000174</td>\n",
       "      <td>living</td>\n",
       "      <td>[alert, attent, mind, intens, concentr, need, demand, peopl, begin, speechread, instruct, eager, progress, understand, everyth, instantli, make, vulner, fatigu, may, strain, forward, muscl, tens, ...</td>\n",
       "      <td>(factory, {'factory': -323.67474030909216, 'living': -332.0057457522294})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -323.67474030909216, 'living': -332.0057457522294}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>165</td>\n",
       "      <td>plant.1000175</td>\n",
       "      <td>living</td>\n",
       "      <td>[need, positiveimag, campaign, posit, imag, sexual, absenc, wing, beat, glassa, bleed, never, endsdesir, endless, corridor, lead, nowher, plant, dont, water, droop, dieyour, collect, erot, imag, c...</td>\n",
       "      <td>(factory, {'factory': -228.119514578958, 'living': -229.96057609900728})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -228.119514578958, 'living': -229.96057609900728}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>167</td>\n",
       "      <td>plant.1000177</td>\n",
       "      <td>living</td>\n",
       "      <td>[consum, associ, say, manufactur, arent, help, situat, label, cloth, dryclean, protect, mani, could, wash, care, breath, fresh, air, housepl, make, home, offic, look, cheeri, accord, research, nas...</td>\n",
       "      <td>(factory, {'factory': -327.79874810699675, 'living': -346.17960393775036})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -327.79874810699675, 'living': -346.17960393775036}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>171</td>\n",
       "      <td>plant.1000182</td>\n",
       "      <td>living</td>\n",
       "      <td>[meant, free, respond, effect, develop, work, got, feel, subject, pastel, use, arrang, place, low, angl, view, present, interest, set, shape, possibl, figur, initi, lay, colour, began, introduc, f...</td>\n",
       "      <td>(factory, {'factory': -574.4030099957007, 'living': -579.7646081606729})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -574.4030099957007, 'living': -579.7646081606729}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>174</td>\n",
       "      <td>plant.1000185</td>\n",
       "      <td>living</td>\n",
       "      <td>[er, one, thing, wont, tell, whether, blaze, hot, day, whether, there, snow, ground, whether, pour, dull, peopl, come, along, there, number, variabl, never, know, think, colin, main, concern, wasn...</td>\n",
       "      <td>(factory, {'factory': -334.99958518457964, 'living': -337.7279757452409})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -334.99958518457964, 'living': -337.7279757452409}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>177</td>\n",
       "      <td>plant.1000188</td>\n",
       "      <td>living</td>\n",
       "      <td>[right, finish, wan, na, look, actual, vertic, distribut, element, ocean, actual, character, three, type, distribut, firstli, substanc, increas, rel, rapidli, depth, becom, constant, group, usual,...</td>\n",
       "      <td>(factory, {'factory': -515.6432778683221, 'living': -537.7984698454885})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -515.6432778683221, 'living': -537.7984698454885}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>178</td>\n",
       "      <td>plant.1000189</td>\n",
       "      <td>living</td>\n",
       "      <td>[get, anoth, speci, next, year, nest, hasnt, nest, know, effort, put, reed, tree, whatev, ive, put, tri, get, nest, nest, thatll, realli, er, thatll, great, thing, er, feed, appl, erm, thought, so...</td>\n",
       "      <td>(factory, {'factory': -2025.2147190174471, 'living': -2081.9296805387885})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -2025.2147190174471, 'living': -2081.9296805387885}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>179</td>\n",
       "      <td>plant.1000190</td>\n",
       "      <td>living</td>\n",
       "      <td>[histori, studi, past, use, document, inscript, evid, historian, record, interpret, event, earliest, day, write, present, day, archaeolog, hand, studi, past, use, object, excav, evid, main, sourc,...</td>\n",
       "      <td>(factory, {'factory': -491.17328381192544, 'living': -506.4500144617781})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -491.17328381192544, 'living': -506.4500144617781}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>181</td>\n",
       "      <td>plant.1000192</td>\n",
       "      <td>living</td>\n",
       "      <td>[gener, bequest, bird, young, chick, need, addit, food, build, flesh, bone, feather, infant, bodi, still, enough, energi, left, break, way, shell, insect, egg, howev, contain, littl, yolk, instead...</td>\n",
       "      <td>(factory, {'factory': -434.1834833688167, 'living': -452.63698896386546})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -434.1834833688167, 'living': -452.63698896386546}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>183</td>\n",
       "      <td>plant.1000195</td>\n",
       "      <td>living</td>\n",
       "      <td>[jacket, detail, glasshous, garden, john, watkin, one, royal, horticultur, societi, collect, conran, octopu, year, publish, theme, seri, fact, seri, build, around, us, gradual, becom, major, part,...</td>\n",
       "      <td>(factory, {'factory': -324.97692638216415, 'living': -332.98244216056054})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -324.97692638216415, 'living': -332.98244216056054}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>184</td>\n",
       "      <td>plant.1000197</td>\n",
       "      <td>living</td>\n",
       "      <td>[power, everyday, languag, scienc, children, encount, school, sometim, introduc, domain, children, littl, everyday, experi, time, involv, work, familiar, phenomena, area, vision, happen, food, eat...</td>\n",
       "      <td>(factory, {'factory': -264.4357919299323, 'living': -280.16386083917655})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -264.4357919299323, 'living': -280.16386083917655}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>186</td>\n",
       "      <td>plant.1000199</td>\n",
       "      <td>living</td>\n",
       "      <td>[enabl, preserv, live, programm, subsequ, retransmiss, sale, countri, telerecord, screen, exactli, domest, televis, screen, except, flat, rather, curv, elimin, distort, imag, captur, film, product...</td>\n",
       "      <td>(factory, {'factory': -386.3417780605314, 'living': -397.8062442973508})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -386.3417780605314, 'living': -397.8062442973508}</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>187</td>\n",
       "      <td>plant.1000200</td>\n",
       "      <td>living</td>\n",
       "      <td>[noth, work, one, everyth, went, wrong, hell, job, even, complet, episod, money, ran, whole, sequenc, effect, rotat, stone, idol, descend, ceil, protrud, spike, move, plant, vine, whole, sequenc, ...</td>\n",
       "      <td>(factory, {'factory': -699.3674057810017, 'living': -728.5016256096594})</td>\n",
       "      <td>factory</td>\n",
       "      <td>{'factory': -699.3674057810017, 'living': -728.5016256096594}</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                id    sense  \\\n",
       "5    plant.1000004   living   \n",
       "8    plant.1000009   living   \n",
       "15   plant.1000016   living   \n",
       "21   plant.1000023   living   \n",
       "30   plant.1000033   living   \n",
       "38   plant.1000041   living   \n",
       "53   plant.1000057   living   \n",
       "75   plant.1000082   living   \n",
       "84   plant.1000091   living   \n",
       "91   plant.1000098   living   \n",
       "93   plant.1000100   living   \n",
       "97   plant.1000104  factory   \n",
       "98   plant.1000105   living   \n",
       "111  plant.1000119   living   \n",
       "120  plant.1000128   living   \n",
       "149  plant.1000157   living   \n",
       "150  plant.1000159   living   \n",
       "153  plant.1000162   living   \n",
       "154  plant.1000163   living   \n",
       "155  plant.1000164   living   \n",
       "156  plant.1000166   living   \n",
       "157  plant.1000167   living   \n",
       "158  plant.1000168   living   \n",
       "163  plant.1000173   living   \n",
       "164  plant.1000174   living   \n",
       "165  plant.1000175   living   \n",
       "167  plant.1000177   living   \n",
       "171  plant.1000182   living   \n",
       "174  plant.1000185   living   \n",
       "177  plant.1000188   living   \n",
       "178  plant.1000189   living   \n",
       "179  plant.1000190   living   \n",
       "181  plant.1000192   living   \n",
       "183  plant.1000195   living   \n",
       "184  plant.1000197   living   \n",
       "186  plant.1000199   living   \n",
       "187  plant.1000200   living   \n",
       "\n",
       "                                                                                                                                                                                                        data  \\\n",
       "5    [breweri, malt, ground, coars, powder, call, grist, thoroughli, mix, larg, vessel, call, mash, tun, hot, pure, water, sugar, dissolv, water, sweet, liquid, call, wort, pump, copper, flower, hop, p...   \n",
       "8    [protech, polyesterbas, film, easili, fit, flat, glass, provid, benefit, toughen, glass, fraction, cost, fit, film, easi, glass, must, thoroughli, clean, made, dustfre, spray, water, ordinari, han...   \n",
       "15   [mani, 60odd, speci, slipper, orchid, best, known, popular, orchid, west, could, extinct, within, year, without, tough, measur, halt, vast, illeg, trade, wild, varieti, dutch, govern, concern, nex...   \n",
       "21   [didnt, know, roll, stone, start, play, river, regular, correspond, rever, minimalist, la, mont, young, introduc, dadaist, movement, fluxu, time, leonard, bernstein, scholarship, berkshir, school,...   \n",
       "30   [ideal, farmer, would, like, grow, larva, egg, new, industri, prawn, hatcheri, grown, meet, demand, hatcheri, taiwan, alon, hatcheri, owner, face, problem, get, suppli, right, feedstuff, metamorph...   \n",
       "38   [move, help, us, combat, trader, put, mani, rare, plant, risk, anoth, action, conserv, endang, plant, netherland, world, centr, plant, trade, second, one, lead, botanist, cite, secretariat, improv...   \n",
       "53                                                                              [helgi, dump, flag, reveal, farm, go, back, guy, real, guid, didnt, insist, simpl, go, direct, plant, flag, regular, interv]   \n",
       "75   [concern, risk, engin, plant, might, acquir, weedi, trait, escap, cultiv, effect, agricultur, environ, could, profound, recent, report, royal, commiss, environment, pollut, conclud, potato, exampl...   \n",
       "84   [instanc, nercfund, project, tortois, southern, franc, local, involv, ten, year, later, tortois, conserv, project, selffund, toast, local, businessmen, embrac, local, peopl, forg, ahead, one, infl...   \n",
       "91   [joseph, gill, john, pinney, harddrink, cousin, live, near, caretak, peggi, marsh, describ, dorothi, one, nicest, girl, ever, saw, arriv, month, help, run, enorm, hous, begin, wordsworth, dorothi,...   \n",
       "93                 [plagu, pestil, famin, prayer, grandfath, smallwe, rememb, occur, litani, book, common, prayer, plagu, pestil, famin, battl, murder, sudden, death, good, lord, deliv, us, bh, plant, sl]   \n",
       "97   [quick, snack, turn, milk, chocol, go, strict, vegan, diet, anim, product, ban, place, tea, coffe, even, fish, egg, salt, sugar, forbidden, like, leap, warm, bed, ici, sea, antarctica, idea, diet,...   \n",
       "98                                                                                                                                                 [voic, peopl, talk, left, music, play, plant, bodi, anim]   \n",
       "111  [nottingham, recogn, sourc, special, contribut, academ, achiev, qualiti, peopl, work, within, interact, enquir, critic, mind, student, staff, pursuit, idea, desir, move, forward, boundari, knowled...   \n",
       "120  [quick, reflex, action, caught, tompaw, short, stout, wooden, handl, steel, claw, set, end, thump, first, deal, day, yanto, work, make, sling, timber, readi, tini, bennion, crane, pick, dump, whar...   \n",
       "149  [shaki, cane, chair, go, tabl, carri, home, head, mum, sat, hour, hour, calligraphi, christma, birthday, card, squar, lush, paper, clean, never, care, interest, wasnt, chore, get, knee, scrub, bru...   \n",
       "150  [shall, see, posit, chang, slightli, encouragingli, sinc, challeng, given, burden, write, book, wide, circl, believ, possibl, consid, biblic, practic, principl, church, plant, enabl, us, promot, p...   \n",
       "153  [mention, domesday, book, inde, evid, seem, indic, one, oldest, largest, settlement, time, domesday, time, armada, reign, elizabeth, mappleton, three, beacon, light, give, warn, invas, villag, mai...   \n",
       "154  [howev, hope, take, time, read, chapter, preced, recip, menu, diet, chart, fail, deepli, impress, mani, way, highfibr, food, help, slim, come, conclus, fplan, major, slim, breakthrough, everyon, s...   \n",
       "155  [light, necessari, viewer, plant, select, dimmest, suffic, interior, decor, also, fall, accept, inhabit, categori, predat, tiger, fish, requir, seclud, hide, place, use, curiowood, dens, growth, g...   \n",
       "156                          [right, float, food, provid, much, koi, vitamin, requir, b, bacteria, probabl, divers, group, live, organ, earth, bacteria, belong, plant, kingdom, rank, lowest, order, speci]   \n",
       "157  [earli, oxygen, level, would, period, depress, order, allow, first, photosynthes, carri, evolv, nevertheless, veget, soon, start, prolifer, liverwort, moss, fern, conifer, plant, flower, appear, q...   \n",
       "158                       [gandhi, recogn, fact, imposs, sustain, one, bodi, without, destruct, bodi, extent, destroy, much, life, think, necessari, susten, preserv, destroy, plant, may, anim, life, food]   \n",
       "163  [aim, reduc, swell, hair, perm, process, mean, hair, structur, protect, long, term, damag, finish, ian, style, sue, hair, use, natur, style, shape, mouss, creat, soft, natur, look, new, style, sha...   \n",
       "164  [alert, attent, mind, intens, concentr, need, demand, peopl, begin, speechread, instruct, eager, progress, understand, everyth, instantli, make, vulner, fatigu, may, strain, forward, muscl, tens, ...   \n",
       "165  [need, positiveimag, campaign, posit, imag, sexual, absenc, wing, beat, glassa, bleed, never, endsdesir, endless, corridor, lead, nowher, plant, dont, water, droop, dieyour, collect, erot, imag, c...   \n",
       "167  [consum, associ, say, manufactur, arent, help, situat, label, cloth, dryclean, protect, mani, could, wash, care, breath, fresh, air, housepl, make, home, offic, look, cheeri, accord, research, nas...   \n",
       "171  [meant, free, respond, effect, develop, work, got, feel, subject, pastel, use, arrang, place, low, angl, view, present, interest, set, shape, possibl, figur, initi, lay, colour, began, introduc, f...   \n",
       "174  [er, one, thing, wont, tell, whether, blaze, hot, day, whether, there, snow, ground, whether, pour, dull, peopl, come, along, there, number, variabl, never, know, think, colin, main, concern, wasn...   \n",
       "177  [right, finish, wan, na, look, actual, vertic, distribut, element, ocean, actual, character, three, type, distribut, firstli, substanc, increas, rel, rapidli, depth, becom, constant, group, usual,...   \n",
       "178  [get, anoth, speci, next, year, nest, hasnt, nest, know, effort, put, reed, tree, whatev, ive, put, tri, get, nest, nest, thatll, realli, er, thatll, great, thing, er, feed, appl, erm, thought, so...   \n",
       "179  [histori, studi, past, use, document, inscript, evid, historian, record, interpret, event, earliest, day, write, present, day, archaeolog, hand, studi, past, use, object, excav, evid, main, sourc,...   \n",
       "181  [gener, bequest, bird, young, chick, need, addit, food, build, flesh, bone, feather, infant, bodi, still, enough, energi, left, break, way, shell, insect, egg, howev, contain, littl, yolk, instead...   \n",
       "183  [jacket, detail, glasshous, garden, john, watkin, one, royal, horticultur, societi, collect, conran, octopu, year, publish, theme, seri, fact, seri, build, around, us, gradual, becom, major, part,...   \n",
       "184  [power, everyday, languag, scienc, children, encount, school, sometim, introduc, domain, children, littl, everyday, experi, time, involv, work, familiar, phenomena, area, vision, happen, food, eat...   \n",
       "186  [enabl, preserv, live, programm, subsequ, retransmiss, sale, countri, telerecord, screen, exactli, domest, televis, screen, except, flat, rather, curv, elimin, distort, imag, captur, film, product...   \n",
       "187  [noth, work, one, everyth, went, wrong, hell, job, even, complet, episod, money, ran, whole, sequenc, effect, rotat, stone, idol, descend, ceil, protrud, spike, move, plant, vine, whole, sequenc, ...   \n",
       "\n",
       "                                                                       func_out  \\\n",
       "5      (factory, {'factory': -291.5626099781813, 'living': -291.6800433457175})   \n",
       "8     (factory, {'factory': -219.21804918526936, 'living': -220.6999942531542})   \n",
       "15     (factory, {'factory': -482.706665346049, 'living': -485.31318249985753})   \n",
       "21    (factory, {'factory': -451.2398071261946, 'living': -453.87001464170635})   \n",
       "30    (factory, {'factory': -299.53540805717313, 'living': -300.2074255847066})   \n",
       "38     (factory, {'factory': -622.0543720341819, 'living': -623.2213303708998})   \n",
       "53   (factory, {'factory': -144.48932220740457, 'living': -144.69310340881904})   \n",
       "75    (factory, {'factory': -345.6128361385043, 'living': -355.15795021642583})   \n",
       "84     (factory, {'factory': -372.9256612396836, 'living': -386.5584717108951})   \n",
       "91        (factory, {'factory': -571.41903872444, 'living': -576.056221753028})   \n",
       "93   (factory, {'factory': -201.55761382256713, 'living': -204.35727865852724})   \n",
       "97    (living, {'factory': -342.44769682361186, 'living': -337.67814820301095})   \n",
       "98    (factory, {'factory': -63.356233554716745, 'living': -63.71509982969176})   \n",
       "111    (factory, {'factory': -605.0678048253524, 'living': -617.5390265614095})   \n",
       "120   (factory, {'factory': -381.2742962105356, 'living': -391.21584447676037})   \n",
       "149   (factory, {'factory': -365.8661031855998, 'living': -369.73000719865075})   \n",
       "150  (factory, {'factory': -327.46794958256646, 'living': -337.23828765893376})   \n",
       "153   (factory, {'factory': -325.0589929979348, 'living': -331.90590264772317})   \n",
       "154  (factory, {'factory': -427.39614901322733, 'living': -442.32821718825267})   \n",
       "155  (factory, {'factory': -248.26358998793822, 'living': -255.52522662956633})   \n",
       "156   (factory, {'factory': -177.59520172795257, 'living': -179.8632370254072})   \n",
       "157  (factory, {'factory': -427.63585758141625, 'living': -438.31243494853356})   \n",
       "158     (factory, {'factory': -171.1696846158845, 'living': -173.833551265479})   \n",
       "163   (factory, {'factory': -457.6589791996005, 'living': -475.84320887674215})   \n",
       "164   (factory, {'factory': -323.67474030909216, 'living': -332.0057457522294})   \n",
       "165    (factory, {'factory': -228.119514578958, 'living': -229.96057609900728})   \n",
       "167  (factory, {'factory': -327.79874810699675, 'living': -346.17960393775036})   \n",
       "171    (factory, {'factory': -574.4030099957007, 'living': -579.7646081606729})   \n",
       "174   (factory, {'factory': -334.99958518457964, 'living': -337.7279757452409})   \n",
       "177    (factory, {'factory': -515.6432778683221, 'living': -537.7984698454885})   \n",
       "178  (factory, {'factory': -2025.2147190174471, 'living': -2081.9296805387885})   \n",
       "179   (factory, {'factory': -491.17328381192544, 'living': -506.4500144617781})   \n",
       "181   (factory, {'factory': -434.1834833688167, 'living': -452.63698896386546})   \n",
       "183  (factory, {'factory': -324.97692638216415, 'living': -332.98244216056054})   \n",
       "184   (factory, {'factory': -264.4357919299323, 'living': -280.16386083917655})   \n",
       "186    (factory, {'factory': -386.3417780605314, 'living': -397.8062442973508})   \n",
       "187    (factory, {'factory': -699.3674057810017, 'living': -728.5016256096594})   \n",
       "\n",
       "    prediction  \\\n",
       "5      factory   \n",
       "8      factory   \n",
       "15     factory   \n",
       "21     factory   \n",
       "30     factory   \n",
       "38     factory   \n",
       "53     factory   \n",
       "75     factory   \n",
       "84     factory   \n",
       "91     factory   \n",
       "93     factory   \n",
       "97      living   \n",
       "98     factory   \n",
       "111    factory   \n",
       "120    factory   \n",
       "149    factory   \n",
       "150    factory   \n",
       "153    factory   \n",
       "154    factory   \n",
       "155    factory   \n",
       "156    factory   \n",
       "157    factory   \n",
       "158    factory   \n",
       "163    factory   \n",
       "164    factory   \n",
       "165    factory   \n",
       "167    factory   \n",
       "171    factory   \n",
       "174    factory   \n",
       "177    factory   \n",
       "178    factory   \n",
       "179    factory   \n",
       "181    factory   \n",
       "183    factory   \n",
       "184    factory   \n",
       "186    factory   \n",
       "187    factory   \n",
       "\n",
       "                                                              scores  \n",
       "5      {'factory': -291.5626099781813, 'living': -291.6800433457175}  \n",
       "8     {'factory': -219.21804918526936, 'living': -220.6999942531542}  \n",
       "15     {'factory': -482.706665346049, 'living': -485.31318249985753}  \n",
       "21    {'factory': -451.2398071261946, 'living': -453.87001464170635}  \n",
       "30    {'factory': -299.53540805717313, 'living': -300.2074255847066}  \n",
       "38     {'factory': -622.0543720341819, 'living': -623.2213303708998}  \n",
       "53   {'factory': -144.48932220740457, 'living': -144.69310340881904}  \n",
       "75    {'factory': -345.6128361385043, 'living': -355.15795021642583}  \n",
       "84     {'factory': -372.9256612396836, 'living': -386.5584717108951}  \n",
       "91        {'factory': -571.41903872444, 'living': -576.056221753028}  \n",
       "93   {'factory': -201.55761382256713, 'living': -204.35727865852724}  \n",
       "97   {'factory': -342.44769682361186, 'living': -337.67814820301095}  \n",
       "98    {'factory': -63.356233554716745, 'living': -63.71509982969176}  \n",
       "111    {'factory': -605.0678048253524, 'living': -617.5390265614095}  \n",
       "120   {'factory': -381.2742962105356, 'living': -391.21584447676037}  \n",
       "149   {'factory': -365.8661031855998, 'living': -369.73000719865075}  \n",
       "150  {'factory': -327.46794958256646, 'living': -337.23828765893376}  \n",
       "153   {'factory': -325.0589929979348, 'living': -331.90590264772317}  \n",
       "154  {'factory': -427.39614901322733, 'living': -442.32821718825267}  \n",
       "155  {'factory': -248.26358998793822, 'living': -255.52522662956633}  \n",
       "156   {'factory': -177.59520172795257, 'living': -179.8632370254072}  \n",
       "157  {'factory': -427.63585758141625, 'living': -438.31243494853356}  \n",
       "158     {'factory': -171.1696846158845, 'living': -173.833551265479}  \n",
       "163   {'factory': -457.6589791996005, 'living': -475.84320887674215}  \n",
       "164   {'factory': -323.67474030909216, 'living': -332.0057457522294}  \n",
       "165    {'factory': -228.119514578958, 'living': -229.96057609900728}  \n",
       "167  {'factory': -327.79874810699675, 'living': -346.17960393775036}  \n",
       "171    {'factory': -574.4030099957007, 'living': -579.7646081606729}  \n",
       "174   {'factory': -334.99958518457964, 'living': -337.7279757452409}  \n",
       "177    {'factory': -515.6432778683221, 'living': -537.7984698454885}  \n",
       "178  {'factory': -2025.2147190174471, 'living': -2081.9296805387885}  \n",
       "179   {'factory': -491.17328381192544, 'living': -506.4500144617781}  \n",
       "181   {'factory': -434.1834833688167, 'living': -452.63698896386546}  \n",
       "183  {'factory': -324.97692638216415, 'living': -332.98244216056054}  \n",
       "184   {'factory': -264.4357919299323, 'living': -280.16386083917655}  \n",
       "186    {'factory': -386.3417780605314, 'living': -397.8062442973508}  \n",
       "187    {'factory': -699.3674057810017, 'living': -728.5016256096594}  "
      ]
     },
     "execution_count": 187,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "outdf[outdf.sense != outdf.prediction].head(40)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Check bass, crane output"
   ]
  }
 ],
 "metadata": {
  "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}