AMR Parsing as Sequence-to-Graph Transduction
This improves AMR parsing for natural language processing by eliminating the need for aligners and external resources, though it is incremental as it builds on existing sequence-to-graph methods.
The paper tackles AMR parsing by proposing an attention-based model that treats it as sequence-to-graph transduction, achieving state-of-the-art SMATCH scores of 76.3% F1 on AMR 2.0 and 70.2% F1 on AMR 1.0.
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).