Event Extraction on ACE2005 using Roberta

1. One CRF layer for trigger detection; 2. Concatenate trigerEmbd, typeEmbd and posEmbd based on the result of trigger word recognition; 3. Use distinct CRF layers for each event type to predict arguments and argument roles.

parameters

python DataLoadAndTrain.py --LOSS_alpha=1 --lr=1e-5 --l2=1e-5 --early_stop=5 --PreTrain_Model="XLMRoberta_large" --batch_size=16

results

[trigger classification] P=0.685 R=0.717 F1=0.701
[argument classification] P=0.413 R=0.256 F1=0.316
[trigger identification] P=0.738 R=0.772 F1=0.755
[argument identification] P=0.457 R=0.283 F1=0.349