Better Combine Them Together! Integrating Syntactic Constituency and Dependency Representations for Semantic Role Labeling

Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 549–559 August 1–6, 2021. ©2021 Association for Computational Linguistics 549 Better Combine Them Together! Integrating Syntactic Constituency and Dependency Representations for Semantic Role Labeling Hao Fei1 and Shengqiong Wu1 and Yafeng Ren2 and Fei Li1 and Donghong Ji1∗ 1. Department of Key Laboratory of Aerospace Information Security and Trusted Computing, Ministry of Education, School of Cyber Science and Engineering, Wuhan University, China 2. Laboratory of Language and Artificial Intelligence, Guangdong University of Foreign Studies, China {hao.fei,whuwsq,renyafeng,lifei csnlp,dhji}@whu.edu.cn Abstract Structural syntax knowledge has been proven effective for semantic role labeling (SRL), while existing works mostly use only one sin- gleton syntax, such as either syntactic depen- dency or constituency tree. In this paper, we explore the integration of heterogeneous syn- tactic representations for SRL. We first con- sider a TreeLSTM-based integration, collabo- ratively learning the phrasal boundaries from the constituency and the semantic relations from dependency. We further introduce a label- aware GCN solution for simultaneously mod- eling the syntactic edges and labels. Experi- mental results demonstrate that by effectively combining the heterogeneous syntactic repre- sentations, our methods yield task improve- ments on both span-based and dependency- based SRL. Also our system achieves new state-of-the-art SRL performances, meanwhile bringing explainable task improvements. 1 Introduction Semantic role labeling (SRL) aims to disclose the predicate-argument structure of a given sentence. Such shallow semantic structures have been shown highly useful for a wide range of downstream tasks in natural language processing (NLP), such as in- formation extraction (Fader et al., 2011; Bastianelli et al., 2013), machine translation (Xiong et al., 2012; Shi et al., 2016) and question answering (Maqsud et al., 2014; Xu et al., 2020). Based on whether to recognize the constituent phrasal span or the syntactic dependency head token of an argu- ment, prior works categorize SRL into two types: the span-based SRL popularized in CoNLL05/12 shared tasks (Carreras and M`arquez, 2005; Prad- han et al., 2013), and the dependency-based SRL introduced in CoNLL08/09 shared tasks (Surdeanu ∗Corresponding author. case det She met her sister in the pub NP VP NP NP S A0 V A1 AM-LOC nmod root ROOT nsubj obj obl PP (1) (2) (3) (4) Figure 1: The mutual benefit to integrate both the (1) syntactic constituency and (4) dependency structures for (2) SRL, based on (3) an example sentence. et al., 2008; Hajiˇc et al., 2009). By adopting var- ious neural network methods, two types of SRL have achieved significant performances in recent years (FitzGerald et al., 2015; He et al., 2017; Fei et al., 2021a) Syntactic features have been extensively verified to be highly effective for SRL (Pradhan et al., 2005; Punyakanok et al., 2008; Marcheggiani and Titov, 2017; Strubell et al., 2018; Zhang et al., 2019). In particular, syntactic dependency features have gained a majority of attention, especially for the dependency-based SRL, considering their close rel- evance with the dependency structure (Roth and Lapata, 2016; He et al., 2018; Xia et al., 2019; Fei et al., …

Content truncated to 500 words