Skip to content

hashbangCoder/Text-Summarization

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Text-Summarization

Text Summarization using Pointer Attention Models

This is a Pytorch implementation of Get To The Point: Summarization with Pointer-Generator Networks by See et. al.

Dependencies

You require :

  1. Pytorch v0.2 with CUDA support
  2. Visdom visualization package for easy monitoring of training progress in web browser
  3. tqdm for terminal-level progress updates

How to get the data

Download the fully-preprocessed data splits here and save yourself the trouble of downloading CoreNLP and subsequent tokenization. This implementation only uses the Finised_Files directory containing train.bin, val.bin, test.bin splits. Move these files to the finished_files directory.

The only other file vocabulary.bin has already been provided in the finished_files in this repo.

Alternatively, follow the instructions here to download and preprocess the dataset. Use the given preprocess.py instead script to generate the data splits as this version does not have Tensorflow dependencies.

How to run

The original hyperparameters as described in the paper are memory intensive, so this implementation uses a smaller RNN hidden_size as the default setting. All other hyperparameters are kept the same. You can trade-off vocabulary_size, batch_size, hidden_size, max_abstract_size and max_article_size to achieve your memory budget.

  1. Fire up a visdom server : python -m visdom.server

  2. To train using default settings, from the repo's root directory : CUDA_VISIBLE_DEVICES=0 python main.py

  3. Monitor the training progress by going to 127.0.0.1:8097 in your web browser or the remote URL is you're executing your code on SSH

Configurations can be changed using command line options.

python main.py --help to get a list of all options.

The model is evaluated periodically during training on a sample from test.binand decoding is done using beam search. The model is also saved after every epoch in Saved-Models

To bootstrap with pre-trained embeddings, you will need to obtain pre-trained Glove/Word2Vec embeddings for words in your vocabulary. OOV words can be assigned a random value. Save this as a Pytorch Tensor embeds.pkl and make sure the size of vocabulary matches size of tensor. The default setting is initialize with random word embeddings since that has been reported to perform better.

About

Text Summarization using Pointer Attention Models

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages