Incorporating Syntax and Semantics in Coreference Resolution with Heterogeneous Graph Attention Network
- 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 underconll_data.
python setup.py install. This will build kernel for extracting top spans implemented using the C++ interface of PyTorch.
python train.py <experiment>- Results are stored in the
log_rootdirectory. - For getting the result of using SpanBERT-Base and SpanBERT-Large model, use
python train.py train_spanbert_base_hgat_dep_srl_two_wayandpython 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.