Entity Commonsense Representation for Neural Abstractive Summarization

This TensorFlow code was used in the experiments of the research paper

Reinald Kim Amplayo^*, Seonjae Lim^* and Seung-won Hwang. Entity Commonsense Representation for Neural Abstractive Summarization. NAACL, 2018.

* Authors have equal contributions

You will need the following data saved in a separate data folder:

• word_vecs.txt: word vectors (we used GloVe vectors which can be downloaded here: http://nlp.stanford.edu/data/glove.840B.300d.zip)
• entity_vecs.txt: entity vectors (we used wiki2vec vectors which can be downloaded here: https://github.com/idio/wiki2vec/raw/master/torrents/enwiki-gensim-word2vec-1000-nostem-10cbow.torrent)
• train.article.txt and valid.article.txt: contains the text to be summarized
• train.title.txt and valid.title.txt: contains the summarized text
• train.entity.txt and valid.entity.txt: contains the entities tagged using the format specified by wiki2vec here: https://github.com/idio/wiki2vec)

To run the code, several parameters are needed to be set in the src/summarization.py. Refer to our paper to determine the recommended values.

To train the model, execute the following code:

python script/train.py

Similarly, to test the model, execute the following code (although test will automatically come after training is done):

python script/test.py

To cite the paper/code, please use this BibTex:

@inproceedings{amplayo2018entity,
	Author = {Reinald Kim Amplayo and Seonjae Lim and Seung-won Hwang},
	Booktitle = {NAACL},
	Location = {New Orleans, LA},
	Year = {2018},
	Title = {Entity Commonsense Representation for Neural Abstractive Summarization},
}

If you have questions, send me an email: rktamplayo at yonsei dot ac dot kr