A Span Selection Model for Semantic Role Labeling
About
We present a simple and accurate span-based model for semantic role labeling (SRL). Our model directly takes into account all possible argument spans and scores them for each label. At decoding time, we greedily select higher scoring labeled spans. One advantage of our model is to allow us to design and use span-level features, that are difficult to use in token-based BIO tagging approaches. Experimental results demonstrate that our ensemble model achieves the state-of-the-art results, 87.4 F1 and 87.0 F1 on the CoNLL-2005 and 2012 datasets, respectively.
Hiroki Ouchi, Hiroyuki Shindo, Yuji Matsumoto• 2018
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Semantic Role Labeling | CoNLL 2005 (WSJ) | F1 Score88.5 | 41 | |
| Span-based Semantic Role Labeling | CoNLL 2005 (Out-of-domain (Brown)) | F1 Score79.6 | 41 | |
| Semantic Role Labeling | CoNLL 2005 (Brown) | F1 Score79.6 | 31 | |
| Semantic Role Labeling | CoNLL 2005 (WSJ (in-domain)) | F1 Score87.6 | 24 | |
| Semantic Role Labeling | CoNLL 2012 | F1 Score87 | 21 | |
| Semantic Role Labeling | CoNLL 2005 (ALL) | F1 Score87.4 | 10 | |
| Semantic Role Labeling | OntoNotes CoNLL-2005 (test) | F1 Score87 | 8 | |
| Span-based Semantic Role Labeling | CoNLL 2005 (In-domain) | F1 Score88.5 | 7 | |
| Span-based Semantic Role Labeling | CoNLL 2012 (In-domain) | F1 Score87 | 6 |
Showing 9 of 9 rows