Dependency Patterns of Complex Sentences and Semantic Disambiguation for Abstract Meaning Representation Parsing

Yuki Yamamoto,Yuji Matsumoto,Taro Watanabe

doi:10.18653/v1/2021.starsem-1.20

Abstract

Abstract Meaning Representation (AMR) is a sentence-level meaning representation based on predicate argument structure. One of the challenges we find in AMR parsing is to capture the structure of complex sentences which expresses the relation between predicates. Knowing the core part of the sentence structure in advance may be beneficial in such a task. In this paper, we present a list of dependency patterns for English complex sentence constructions designed for AMR parsing. With a dedicated pattern matcher, all occurrences of complex sentence constructions are retrieved from an input sentence. While some of the subordinators have semantic ambiguities, we deal with this problem through training classification models on data derived from AMR and Wikipedia corpus, establishing a new baseline for future works. The developed complex sentence patterns and the corresponding AMR descriptions will be made public.

Highlights

Meaning Representation (AMR) is a sentence-level meaning representation based on predicate argument structure (Banarescu et al, 2013)
With the motivation to aid Abstract Meaning Representation (AMR) parsing task, we present a method to retrieve all occurrences of complex sentence constructions from an input sentence using a dedicated pattern matcher
With the support of weakly supervised data derived from Wikipedia, we achieve the scores of 75.65% and 83.94% on macro and micro F1 respectively, establishing a new baseline for coherence relation classification of complex sentence constructions in AMR framework

Summary

Introduction

Abstract Meaning Representation (AMR) is a sentence-level meaning representation based on predicate argument structure (Banarescu et al, 2013). While early studies (Flanigan et al, 2014; Wang et al, 2015; Artzi et al, 2015; Pust et al, 2015) used dependency parsers to integrate syntactic features to their models, recent deep neural network-based approaches (Konstas et al, 2017; Peng et al, 2017; Zhang et al, 2019; Cai and Lam, 2020) tend to encode the input sentence as a sequence without considering its syntactic structure. Syntactic and semantic structures share much in common. It is assumed QVXEM 6 ORQJ DGYPRG DGYFO DV PDUN $5* 6 DVORQJDV RS

Methods

Results

Conclusion