Multi-word Expression and Supersense Tagging on DiMSUM 1.0 (test)
73.5MWE PrecisionKirilin et al.
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Kirilin et al.2020.04 | 73.5 | 48.4 | 58.4 | 56.8 | 59.2 | 58 | 85.3 | 59 | 57.2 | 58.1 | |
| neural CRF taggerEmbeddings=BERT + GloVe, POS/lemma constraints=None, Evaluation=Zero-shot2020.04 | 53 | 49.2 | 51 | 50.8 | 55.1 | 52.9 | 76.5 | 51.2 | 53.9 | 52.5 | |
| neural CRF taggerEmbeddings=BERT + GloVe, POS/lemma constraints=Predicted, Evaluation=Zero-shot2020.04 | 48.8 | 50.7 | 49.7 | 49.1 | 53.9 | 51.4 | 75.1 | 49.1 | 53.3 | 51.1 | |
| neural CRF taggerEmbeddings=BERT + GloVe, POS/lemma constraints=Gold, Evaluation=Zero-shot2020.04 | 47.9 | 52.2 | 50 | 52.1 | 56.5 | 54.2 | 76.9 | 51.3 | 55.7 | 53.4 |