The Role of Reentrancies in Abstract Meaning Representation Parsing

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 2198–2207 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 2198 The Role of Reentrancies in Abstract Meaning Representation Parsing Ida Szubert1∗, Marco Damonte2∗†, Shay B. Cohen1, Mark Steedman1 1 University of Edinburgh, 2 Amazon Alexa, Cambridge UK k.i.szubert@sms.ed.ac.uk , dammarco@amazon.com , scohen@inf.ed.ac.uk , steedman@inf.ed.ac.uk Abstract Abstract Meaning Representation (AMR) parsing aims at converting sentences into AMR representations. These are graphs and not trees because AMR supports reentrancies (nodes with more than one parent). Following previous findings on the importance of reen- trancies for AMR, we empirically find and discuss several linguistic phenomena respon- sible for reentrancies in AMR, some of which have not received attention before. We cate- gorize the types of errors AMR parsers make with respect to reentrancies. Furthermore, we find that correcting these errors provides an in- crease of up to 5% Smatch in parsing perfor- mance and 20% in reentrancy prediction. 1 Introduction Abstract Meaning Representation (AMR) is a se- mantic formalism used to annotate natural lan- guage sentences as graphs. The task of AMR pars- ing is to convert sentences into AMR graphs (Ba- narescu et al., 2013) — rooted and directed acyclic graphs where nodes represent concepts and edges represent semantic relations between them. The AMR for the sentence I want you to believe me is shown in Figure 1. One of the main properties of AMR, and the reason why sentences are represented as graphs rather than trees, is the presence of nodes with multiple parents, called reentrancies, as demon- strated in Figure 1, where the node I has two parents. Reentrancies complicate AMR parsing and require the addition of specific transitions in transition-based parsing (Wang et al., 2015; Da- monte et al., 2017) or of pre- and post-processing steps in sequence-to-sequence parsing (van Noord and Bos, 2017). Enabling AMR parsers to predict ∗ Equal contribution † Work done while at University of Edinburgh want-01 I you believe-01 :ARG0 :ARG1 :ARG0 :ARG1 Figure 1: AMR for I want you to believe me. reentrancy structures correctly is of particular im- portance because it separates AMR parsing from semantic parsing based on tree structures (Steed- man, 2000; Liang, 2013; Cheng et al., 2017). Reentrancy is however not an AMR-specific prob- lem (Kuhlmann and Jonsson, 2015), and other for- malisms can benefit from a better understanding of how to parse such structures. Nevertheless, to our knowledge, the AMR literature lacks any detailed discussion of the types and linguistic causes of re- entrant structures. We aim to fill the gap by de- scribing the phenomena causing reentrancies and quantifying their prevalence in the AMR corpus. We identify sources of reentrancy which have not been acknowledged in the AMR literature such as adjunct control, verbalization, and pragmatics. AMR parsers are evaluated using Smatch (Cai and Knight, 2013), which however does not ex- plicitly assess the parsers’ ability to recover reen- trancies. Damonte et al. (2017) introduced a mea- sure of reentrancy prediction, which computes the Smatch …

Content truncated to 500 words