This repository contains code for training and running the neural coreference models decribed in two papers:

• [Coming Soon] "Deep Reinforcement Learning for Mention-Ranking Coreference Models", Kevin Clark and Christopher D. Manning, EMNLP 2016.
• "Improving Coreference Resolution by Learning Entity-Level Distributed Representations", Kevin Clark and Christopher D. Manning, ACL 2016.

Theano, numpy, and scikit-learn. It also uses a slightly modified version of keras 0.2; run python setup.py install in the modified_keras directory to install.

The easiest way of doing this is within Stanford's CoreNLP, where forward-prop for the models has been implemented in Java. Example usage:

java -Xmx5g -cp stanford-corenlp.jar edu.stanford.nlp.pipeline.StanfordCoreNLP -annotators tokenize,ssplit,pos,lemma,ner,parse,mention,coref -coref.algorithm neural -file example_file.txt

You will need to fork the latest version from github and download the latest models from here.

1. Download pretrained word embeddings. We use 50 dimensional word2vec embeddings for English (link) and 64 dimenensional polyglot embeddings for Chinese (link) in our paper.
2. Run the NeuralCorefDataExporter class in the development version Stanford's CoreNLP using this properties file. This does mention detection and feature extraction on the CoNLL data and then outputs the results as json.
3. Run run_all.py, preferably on a GPU. Training takes roughly 4 days on a GTX TITAN GPU.