batcher.py

# Copyright 2016 The TensorFlow Authors. All Rights Reserved.
# Modifications Copyright 2017 Abigail See
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================

"""This file contains code to process data into batches"""

import queue
from random import shuffle
from threading import Thread
import time
import numpy as np
import tensorflow as tf
import data
from nltk.tokenize import sent_tokenize
import glob
import codecs
import json
FLAGS = tf.app.flags.FLAGS
class Example(object):
  """Class representing a train/val/test example for text summarization."""

  def __init__(self, review, vocab, hps, input=None):
    """Initializes the Example, performing tokenization and truncation to produce the encoder, decoder and target sequences, which are stored in self.

    Args:
      article: source text; a string. each token is separated by a single space.
      abstract_sentences: list of strings, one per abstract sentence. In each sentence, each token is separated by a single space.
      vocab: Vocabulary object
      hps: hyperparameters
    """
    self.hps = hps

    # Get ids of special tokens
    start_decoding = vocab.word2id(data.START_DECODING)
    stop_decoding = vocab.word2id(data.STOP_DECODING)
    stop_doc = vocab.word2id(data.STOP_DECODING_DOCUMENT)




    #abstract_words = review.split() # list of strings
    #abs_ids = [vocab.word2id(w) for w in abstract_words] # list of word ids; OOVs are represented by the id for UNK token

    if input !=None:
        review_sentence = sent_tokenize(review)

        article = input
        article_words = article.split()  # list of strings
        if len(article_words) > hps.max_enc_steps:
            article_words = article_words[:hps.max_enc_steps]
        self.enc_len = len(article_words)  # store the length after truncation but before padding
        self.enc_input = [vocab.word2id(w) for w in
                          article_words]  # list of word ids; OOVs are represented by the id for UNK token
        self.original_review_input =input
        self.original_review_output = review




        abstract_sentences = [x.strip() for x in review_sentence]
        abstract_words = []
        for i in range(len(abstract_sentences)):
            if i >= hps.max_dec_sen_num:
                abstract_words = abstract_words[:hps.max_dec_sen_num]
                break
            abstract_sen = abstract_sentences[i]
            abstract_sen_words = abstract_sen.split()
            if len(abstract_sen_words) > hps.max_dec_steps:
                abstract_sen_words = abstract_sen_words[:hps.max_dec_steps]
            abstract_words.append(abstract_sen_words)

        if len(abstract_words[-1]) < hps.max_dec_steps:
            abstract_words[-1].append(stop_doc)
    else:


        review_sentence = sent_tokenize(review)

        article = review_sentence[0]
        article_words = article.split()  # list of strings
        if len(article_words) > hps.max_enc_steps:

            article_words = article_words[:hps.max_enc_steps]
        self.enc_len = len(article_words)  # store the length after truncation but before padding
        self.enc_input = [vocab.word2id(w) for w in
                          article_words]  # list of word ids; OOVs are represented by the id for UNK token
        self.original_review_input = review_sentence[0]
        self.original_review_output = " ".join(review_sentence[1:])

        review_sentence = review_sentence[1:]




        abstract_sentences = [x.strip() for x in review_sentence]
        abstract_words = []
        for i in range(len(abstract_sentences)):
            if i >= hps.max_dec_sen_num:
                abstract_words = abstract_words[:hps.max_dec_sen_num]
                break
            abstract_sen = abstract_sentences[i]
            abstract_sen_words = abstract_sen.split()
            if len(abstract_sen_words) > hps.max_dec_steps:
                abstract_sen_words = abstract_sen_words[:hps.max_dec_steps]
            abstract_words.append(abstract_sen_words)

        if len(abstract_words[-1]) < hps.max_dec_steps:
            abstract_words[-1].append(stop_doc)
        '''if len(abstract_sentences) < hps.max_dec_sen_num:
            abstract_words.append([stop_doc])'''



    # abstract_words = abstract.split() # list of strings
    abs_ids = [[vocab.word2id(w) for w in sen] for sen in
               abstract_words]  # list of word ids; OOVs are represented by the id for UNK token

    # Get the decoder input sequence and target sequence
    self.dec_input, self.target = self.get_dec_inp_targ_seqs(abs_ids, hps.max_dec_sen_num, hps.max_dec_steps,
                                                              start_decoding,stop_decoding)  # max_sen_num,max_len, start_doc_id, end_doc_id,start_id, stop_id
    self.dec_len = len(self.dec_input)
    self.dec_sen_len = [len(sentence) for sentence in self.target]
    self.original_review = review



  def get_dec_inp_targ_seqs(self, sequence, max_sen_num,max_len, start_id, stop_id):
    """Given the reference summary as a sequence of tokens, return the input sequence for the decoder, and the target sequence which we will use to calculate loss. The sequence will be truncated if it is longer than max_len. The input sequence must start with the start_id and the target sequence must end with the stop_id (but not if it's been truncated).

    Args:
      sequence: List of ids (integers)
      max_len: integer
      start_id: integer
      stop_id: integer

    Returns:
      inp: sequence length <=max_len starting with start_id
      target: sequence same length as input, ending with stop_id only if there was no truncation
    """

    inps = sequence[:]
    targets = sequence[:]

    if len(inps) > max_sen_num:
        inps = inps[:max_sen_num]
        targets = targets[:max_sen_num]

    for i in range(len(inps)):
        inps[i] = [start_id] + inps[i][:]
        if len(inps[i]) > max_len:
            inps[i] = inps[i][:max_len]

    for i in range(len(targets)):
        if len(targets[i]) >= max_len:
            targets[i] = targets[i][:max_len - 1]  # no end_token
            targets[i].append(stop_id)  # end token
        else:
            targets[i]=targets[i] +[stop_id]

    return inps, targets

  def pad_decoder_inp_targ(self, max_sen_len, max_sen_num, pad_doc_id):
      """Pad decoder input and target sequences with pad_id up to max_len."""

      while len(self.dec_sen_len) < max_sen_num:
          self.dec_sen_len.append(1)

      for i in range(len(self.dec_input)):
          while len(self.dec_input[i]) < max_sen_len:
              self.dec_input[i].append(pad_doc_id)

      while len(self.dec_input) < max_sen_num:
          self.dec_input.append([pad_doc_id for i in range(max_sen_len)])

      for i in range(len(self.target)):
          while len(self.target[i]) < max_sen_len:
              self.target[i].append(pad_doc_id)

      while len(self.target) < max_sen_num:
          self.target.append([pad_doc_id for i in range(max_sen_len)])
          # print (self.target)

  def pad_encoder_input(self, max_len, pad_id):
    """Pad the encoder input sequence with pad_id up to max_len."""
    while len(self.enc_input) < max_len:
      self.enc_input.append(pad_id)



class Batch(object):
  """Class representing a minibatch of train/val/test examples for text summarization."""

  def __init__(self, example_list, hps, vocab):
    """Turns the example_list into a Batch object.

    Args:
       example_list: List of Example objects
       hps: hyperparameters
       vocab: Vocabulary object
    """
    self.pad_id = vocab.word2id(data.PAD_TOKEN) # id of the PAD token used to pad sequences
    if FLAGS.run_method == 'auto-encoder':
        self.init_encoder_seq(example_list, hps)  # initialize the input to the encoder
    self.init_decoder_seq(example_list, hps) # initialize the input and targets for the decoder
    self.store_orig_strings(example_list) # store the original strings



  def init_encoder_seq(self, example_list, hps):

    #print ([ex.enc_len for ex in example_list])

    max_enc_seq_len = max([ex.enc_len for ex in example_list])

    # Pad the encoder input sequences up to the length of the longest sequence
    for ex in example_list:
      ex.pad_encoder_input(max_enc_seq_len, self.pad_id)

    # Initialize the numpy arrays
    # Note: our enc_batch can have different length (second dimension) for each batch because we use dynamic_rnn for the encoder.
    self.enc_batch = np.zeros((hps.batch_size, max_enc_seq_len), dtype=np.int32)
    self.enc_lens = np.zeros((hps.batch_size), dtype=np.int32)
    #self.enc_padding_mask = np.zeros((hps.batch_size, max_enc_seq_len), dtype=np.float32)

    # Fill in the numpy arrays
    for i, ex in enumerate(example_list):
      #print (ex.enc_input)
      self.enc_batch[i, :] = ex.enc_input[:]
      self.enc_lens[i] = ex.enc_len
      '''for j in range(ex.enc_len):
        self.enc_padding_mask[i][j] = 1'''




  def init_decoder_seq(self, example_list, hps):

    for ex in example_list:
      ex.pad_decoder_inp_targ(hps.max_dec_steps, hps.max_dec_sen_num,self.pad_id)

    # Initialize the numpy arrays.
    # Note: our decoder inputs and targets must be the same length for each batch (second dimension = max_dec_steps) because we do not use a dynamic_rnn for decoding. However I believe this is possible, or will soon be possible, with Tensorflow 1.0, in which case it may be best to upgrade to that.
    self.dec_batch = np.zeros((hps.batch_size, hps.max_dec_sen_num, hps.max_dec_steps), dtype=np.int32)
    self.target_batch = np.zeros((hps.batch_size, hps.max_dec_sen_num, hps.max_dec_steps), dtype=np.int32)
    self.dec_padding_mask = np.zeros((hps.batch_size* hps.max_dec_sen_num, hps.max_dec_steps),
                                     dtype=np.float32)
    self.dec_sen_lens = np.zeros((hps.batch_size, hps.max_dec_sen_num), dtype=np.int32)
    self.dec_lens = np.zeros((hps.batch_size), dtype=np.int32)

    for i, ex in enumerate(example_list):
        self.dec_lens[i] = ex.dec_len
        self.dec_batch[i, :, :] = np.array(ex.dec_input)
        self.target_batch[i] = np.array(ex.target)
        for j in range(len(ex.dec_sen_len)):
            self.dec_sen_lens[i][j] = ex.dec_sen_len[j]


    self.target_batch = np.reshape(self.target_batch,
                                   [hps.batch_size*hps.max_dec_sen_num, hps.max_dec_steps])

    for j in range(len(self.target_batch)):
        for k in range(len(self.target_batch[j])):
            if int(self.target_batch[j][k]) != self.pad_id:
                self.dec_padding_mask[j][k] = 1
    #self.dec_padding_mask = np.reshape(self.dec_padding_mask, [hps.batch_size*hps.max_dec_sen_num, hps.max_dec_steps])

  def store_orig_strings(self, example_list):
    """Store the original article and abstract strings in the Batch object"""

    self.original_review_output = [ex.original_review_output for ex in example_list] # list of lists
    if FLAGS.run_method == 'auto-encoder':
        self.original_review_inputs = [ex.original_review_input for ex in example_list]  # list of lists



class GenBatcher(object):

    def __init__(self, vocab, hps):
        self._vocab = vocab
        self._hps = hps

        self.train_queue = self.fill_example_queue("review_generation_dataset/train/*", mode ="train")
        self.test_queue = self.fill_example_queue("review_generation_dataset/test/*",  mode ="test")
        #self.test_queue = self.fill_example_queue("/home/xujingjing/code/review_summary/dataset/review_generation_dataset/test/*")
        self.train_batch = self.create_batch(mode="train")
        self.test_batch = self.create_batch(mode="test", shuffleis=False)
        #train_batch = self.create_bach(mode="train")

    def create_batch(self, mode="train", shuffleis=True):
        all_batch = []

        if mode == "train":
            num_batches = int(len(self.train_queue) / self._hps.batch_size)
            if shuffleis:
                shuffle(self.train_queue)
        elif mode == 'test':
            num_batches = int(len(self.test_queue) / self._hps.batch_size)

        for i in range(0, num_batches):
            batch = []
            if mode == 'train':
                batch += (self.train_queue[i * self._hps.batch_size:i * self._hps.batch_size + self._hps.batch_size])
            elif mode == 'test':
                batch += (self.test_queue[i * self._hps.batch_size:i * self._hps.batch_size + self._hps.batch_size])

            all_batch.append(Batch(batch, self._hps, self._vocab))
        return all_batch


    def get_batches(self, mode="train"):


        if mode == "train":
            shuffle(self.train_batch)
            return self.train_batch
        elif mode == 'test':
            return self.test_batch




    def fill_example_queue(self, data_path, mode = "test"):

        new_queue =[]

        filelist = glob.glob(data_path)  # get the list of datafiles
        assert filelist, ('Error: Empty filelist at %s' % data_path)  # check filelist isn't empty
        filelist = sorted(filelist)
        if mode == "train":
            filelist = filelist

        for f in filelist:


            reader = codecs.open(f, 'r', 'utf-8')
            while True:
                string_ = reader.readline()
                if not string_: break
                dict_example = json.loads(string_)
                review = dict_example["review"]
                if(len(sent_tokenize(review))<2):
                    continue
                example = Example(review, self._vocab, self._hps)
                new_queue.append(example)
        return new_queue