This repository contain relevant code (in form of .ipynb) to perform the task of pronoun resolution using end-to-end neural co-reference resolution system. Our implementation makes use of BERT based embeddings to generate vector representation for spans of text. It then uses a FFNN to ranking candidate antecedent spans for the given pronoun.
The current code performs pronoun resolution on the GAP Dataset.
Start up a IPython server instance to load the file and execute all cells. You can download the trained model from here and load it to test the co-reference system.