Incorporating Syntax and Semantics in Coreference Resolution with Heterogeneous Graph Attention Network

Setup

• pip install -r requirements.txt
• ./setup_training.sh <ontonotes/path/ontonotes-release-5.0> conll_data. This assumes that you have access to OntoNotes 5.0. The preprocessed data will be included under conll_data.

Build Kernels

• python setup.py install. This will build kernel for extracting top spans implemented using the C++ interface of PyTorch.

Training

• python train.py <experiment>
• Results are stored in the log_root directory.
• For getting the result of using SpanBERT-Base and SpanBERT-Large model, use python train.py train_spanbert_base_hgat_dep_srl_two_way and python train.py train_spanbert_large_hgat_dep_srl_two_way
• Finetuning a SpanBERT large model on OntoNotes requires access to a 32GB GPU, while the base model can be trained in a 16GB GPU.