Free the Plural: Unrestricted Split-Antecedent Anaphora Resolution

Introduction

This repository contains code introduced in the following paper:

Free the Plural: Unrestricted Split-Antecedent Anaphora Resolution
Juntao Yu, Nafise Sadat Moosavi, Silviu Paun and Massimo Poesio
In Proceedings of he 28th International Conference on Computational Linguistics (COLING), 2020

Setup Environments

• The code is written in Python 2, the compatibility to Python 3 is not guaranteed.
• Before starting, you need to install all the required packages listed in the requirment.txt using pip install -r requirements.txt.
• After that modify and run extract_bert_features/extract_bert_features.sh to compute the BERT embeddings for your training or testing.
• You also need to download context-independent word embeddings such as fasttext or GloVe embeddings that required by the system.

To use a pre-trained model

• Pre-trained models can be download from this link. We provide the best model for reported in our paper.

• Choose the model you want to use and copy them to the logs/ folder.

• Modifiy the test_path accordingly in the experiments.conf:

  • the test_path is the path to .jsonlines file, each line of the .jsonlines file is a batch of sentences and must in the following format:

  {
  "clusters": [[0, 4],[1],[2],[3],], #Coreference use the indices of the mention
  "mentions": [[0,0],[2,3],[5,5],[7,8],[10,10],[12,13]], #mentions [start_index, end_index]
  "plurals": [[5,1],[5,3]], #plural [anaphor, antecedent] pairs, "both cars" --> "a car", "another car"
  "doc_key": "nw",
  "sentences": [["John", "has", "a", "car", "."], ["Mary", "has", "another", "car", "."] ["John", "washed", "both", "cars", "yesteday","."]]
  }
  

  • The mentions only contain two properties [start_index, end_index] the indices are counted in document level and both inclusive.
  • For coreference clusters (includes singleton clusters) are represented by their mention indices.
  • For plural pairs, each pair contains two mention indices the first one is the anaphora and the second one is the antecedent.

• Then use python evaluate.py config_name to start your evaluation

To train your own model

• You will need additionally to create the character vocabulary by using python get_char_vocab.py train.jsonlines dev.jsonlines
• Then you can start training by using python train.py config_name