the code of 《CAN WE GET MORE FROM LABELS? EVENT EXTRACTION WITHOUT TRIGGERS》IEEE ICASSP 2022
First of all you need to have the original data of ACE2005. Then refer to https://github.com/nlpcl-lab/ace2005-preprocessing for preprocessing. Then put the output file into SLE-DA/data to train the model
python run_args.py --do_train
to eval the model
python run_args.py --do_eval
results are shown at below table,ours(SLE-DA)
| models | Event Detection F1 | Argument Extraction F1 |
|---|---|---|
| DMCNN | 69.1 | 53.5 |
| PLMEE | 80.7 | 58.0 |
| JRNN | 69.3 | 55.4 |
| JMEE | 73.7 | 60.3 |
| BERT-QA | 72.4 | 53.5 |
| MQAEE | 73.8 | 55.0 |
| TBNNAM | 69.9 | - |
| SLE-DA | 77.4 | 58.7 |