A Differentiable Relaxation of Graph Segmentation and Alignment for AMR Parsing

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 9075–9091 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 9075 A Differentiable Relaxation of Graph Segmentation and Alignment for AMR Parsing Chunchuan Lyu1 Shay B. Cohen1 Ivan Titov1,2 1 ILCC, School of Informatics, University of Edinburgh 2 ILLC, University of Amsterdam chunchuan.lv@gmail.com scohen@inf.ed.ac.uk ititov@inf.ed.ac.uk Abstract Abstract Meaning Representations (AMR) are a broad-coverage semantic formalism which represents sentence meaning as a directed acyclic graph. To train most AMR parsers, one needs to segment the graph into subgraphs and align each such subgraph to a word in a sen- tence; this is normally done at preprocessing, relying on hand-crafted rules. In contrast, we treat both alignment and segmentation as la- tent variables in our model and induce them as part of end-to-end training. As marginalizing over the structured latent variables is infeasi- ble, we use the variational autoencoding frame- work. To ensure end-to-end differentiable op- timization, we introduce a differentiable relax- ation of the segmentation and alignment prob- lems. We observe that inducing segmentation yields substantial gains over using a ‘greedy’ segmentation heuristic. The performance of our method also approaches that of a model that relies on the segmentation rules of Lyu and Titov (2018), which were hand-crafted to handle individual AMR constructions. 1 Introduction Abstract Meaning Representation (AMR; Ba- narescu et al. 2013) is a broad-coverage semantic formalism which represents sentence meaning as rooted labeled directed acyclic graphs. The rep- resentations have been exploited in a wide range of tasks, including text summarization (Liu et al., 2015; Dohare and Karnick, 2017; Hardy and Vla- chos, 2018), machine translation (Jones et al., 2012; Song et al., 2019), paraphrase detection (Issa et al., 2018) and question answering (Mitra and Baral, 2016). An AMR graph can be regarded as consisting of multiple concept subgraphs, which can be indi- vidually aligned to sentence tokens (Flanigan et al., 2014). In Figure 1, each dashed box represents the boundary of a single semantic subgraph. Red arrows represent the alignment between subgraphs 2 The opinion of the boy boy ARG0 opine-01 thing ARG1 0 1 2 0 1 3 4 Figure 1: An example of AMR, the dashed red arrows mark latent alignment. Dashed blue boxes represent the latent segmentation. and tokens. For example, ‘(o / opine-01: ARG1 (t / thing))’ refers to a combination of the predicate ‘opine-01’ and a filler of its semantic role ARG1. Intuitively, this subgraph needs to be aligned to the token ‘opinion’. Similarly, ‘(b / boy)’ should be aligned to the token ‘boy’. Given such an alignment and segmentation, it is straightforward to construct a simple parser: parsing can be framed as tagging input tokens with subgraphs (including empty sub- graphs), followed by predicting relations between the subgraphs. The key obstacle to training such an AMR parser is that the segmentation and alignment between AMR subgraphs and words are latent, i.e. not annotated in the data. Most previous work adopts a pipeline approach to handling this obstacle. They rely …

Content truncated to 500 words