This repository provides the source code of our paper "A Hybrid Detection and Generation Framework with Separate Encoders for Event Extraction". (Received by EACL)
Currently we provide a Baseline version of our implementation based on GTEE-DYNPREF, and the code used in our paper will be coming soon after the paper is finalized for publication.
All datasets can be placed in data or elsewhere.
prepreocessing/process_dygiepp.py from OneIE v0.4.8
preprocessing/process_ace.py from OneIE v0.4.8
prepreocessing/process_ere.py from OneIE v0.4.8
To train baseline on different datasets, specify the configurations located in scripts/*.sh
bash scripts/train_base_generator.sh and bash scripts/train_base_IC.sh can be executed in parallel, but bash scripts/eval_base_IC.sh needs to be executed first to get the file az_file used for bash scripts/eval_base_generator.sh