IrisCHEN04/hedge_detection_distill_bert

Notice

Please be careful when using the model. As a linguist, I do not think it meet my standards. But it may be OK for you if you are not from my field and do not need to be that prudent on the concept.

Train

I trained it for 5 epochs using the below dataset:

Farkas, R., Vincze, V., Móra, G., Csirik, J., & Szarvas, G. (2010). The CoNLL-2010 shared task: Learning to detect hedges and their scope in natural language text. In Proceedings of the Fourteenth Conference on Computational Natural Language Learning—Shared Task (pp. 1-12). Association for Computational Linguistics.

Training Metrics

Metric	Value
Global Step	2,895
Training Loss	0.0056
Epochs	5.0
Training Runtime	110.40 seconds
Train Samples/Second	419.40
Train Steps/Second	26.22
Total FLOPs	9.53 × 10¹⁴

Training History

Epoch	Training Loss	Validation Loss	Precision	Recall	F1 Score	Accuracy
1	0.0181	0.0075	0.8882	0.8773	0.8827	0.9975
2	0.0053	0.0070	0.8711	0.9018	0.8862	0.9975
3	0.0044	0.0076	0.9141	0.8972	0.9056	0.9980
4	0.0024	0.0089	0.8880	0.9003	0.8941	0.9977
5	0.0018	0.0100	0.8965	0.9034	0.8999	0.9978