AMR Parsing as Sequence-to-Graph Transduction

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 80–94 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 80 AMR Parsing as Sequence-to-Graph Transduction Sheng Zhang Xutai Ma Kevin Duh Johns Hopkins UniversityBenjamin Van Durme {zsheng2, xutai ma}@jhu.edu {kevinduh, vandurme}@cs.jhu.edu Abstract 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). 1 Introduction Abstract Meaning Representation (AMR, Ba- narescu et al., 2013) parsing is the task of trans- ducing natural language text into AMR, a graph- based formalism used for capturing sentence-level semantics. Challenges in AMR parsing include: (1) its property of reentrancy – the same concept can participate in multiple relations – which leads to graphs in contrast to trees (Wang et al., 2015); (2) the lack of gold alignments between nodes (concepts) in the graph and words in the text which limits attempts to rely on explicit alignments to generate training data (Flanigan et al., 2014; Wang et al., 2015; Damonte et al., 2017; Foland and Mar- tin, 2017; Peng et al., 2017b; Groschwitz et al., 2018; Guo and Lu, 2018); and (3) relatively lim- ited amounts of labeled data (Konstas et al., 2017). Recent attempts to overcome these challenges include: modeling alignments as latent vari- ables (Lyu and Titov, 2018); leveraging exter- nal semantic resources (Artzi et al., 2015; Bjerva et al., 2016); data augmentation (Konstas et al., 2017; van Noord and Bos, 2017b); and employ- ing attention-based sequence-to-sequence mod- els (Barzdins and Gosko, 2016; Konstas et al., 2017; van Noord and Bos, 2017b). In this paper, we introduce a different way to handle reentrancy, and propose an attention-based possible-01 help-01 victim ARG1 ARG1 ARG0 possible-01 help-01 2 victim 3 victim 3 ARG1 ARG1 ARG0 1 (a) (b) Figure 1: Two views of reentrancy in AMR for an ex- ample sentence “The victim could help himself.” (a) A standard AMR graph. (b) An AMR tree with node indices as an extra layer of annotation, where the cor- responding graph can be recovered by merging nodes of the same index and unioning their incoming edges. model that treats AMR parsing as sequence-to- graph transduction. The proposed model, sup- ported by an extended pointer-generator network, is aligner-free and can be effectively trained with limited amount of labeled AMR data. Exper- iments on two publicly available AMR bench- marks demonstrate that our parser clearly outper- forms the previous best parsers on both bench- marks. It achieves the best reported SMATCH scores: 76.3% F1 on LDC2017T10 and 70.2% F1 on LDC2014T12. We also provide extensive ab- lative and qualitative studies, quantifying the con- tributions from each component. Our model im- plementation is available …

Content truncated to 500 words