Sheng Zhang, Xutai Ma, Kevin Duh, Benjamin Van Durme
We propose an attention-based model that treats AMR parsing as sequence-to-graph transduction. Unlike most AMR parsers that rely on pre-trained aligners, external semantic resources, or data augmentation, our proposed parser is aligner-free, and it can be effectively trained with limited amounts of labeled AMR data. Our experimental results outperform all previously reported SMATCH scores, on both AMR 2.0 (76.3% F1 on LDC2017T10) and AMR 1.0 (70.2% F1 on LDC2014T12).
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Semantic Parsing | LDC2014T12 | F1 Full | 70.2 | Two-stage Sequence-to-Graph Transducer |
| Semantic Parsing | LDC2017T10 | Smatch | 76.3 | Sequence-to-Graph Transduction |
| Semantic Parsing | LDC2014T12: | F1 Full | 0.7 | Sequence-to-Graph Transduction |
| Semantic Parsing | LDC2014T12: | F1 Newswire | 0.75 | Sequence-to-Graph Transduction |
| AMR Parsing | LDC2014T12 | F1 Full | 70.2 | Two-stage Sequence-to-Graph Transducer |
| AMR Parsing | LDC2017T10 | Smatch | 76.3 | Sequence-to-Graph Transduction |
| AMR Parsing | LDC2014T12: | F1 Full | 0.7 | Sequence-to-Graph Transduction |
| AMR Parsing | LDC2014T12: | F1 Newswire | 0.75 | Sequence-to-Graph Transduction |