/NAACL2021-Event-Time-Extraction-and-Propagation-via-Graph-Attention-Networks

Primary LanguagePython

Event Time Extraction and Propagation via Graph Attention Networks

Resources for paper "Event Time Extraction and Propagation via Graph Attention Networks". NAACL 2021

Data

The 4-tuple annotations are in data.json. The data are at event level since the time representations for all coreferential mentions should be consistent.

Because of the license restriction, we are not able to provide pre-processed ACE2005 data (doc_ace_entities.json, doc_ace.json).

Dependencies

  • torch=1.2.0
  • tqdm=4.36.1
  • transformers=3.3.0
  • numpy=1.18.4
  • dgl-cuda10.0=0.4.3post2