Syntax-driven Approach for Semantic Role Labeling

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 7129–7139 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 7129 Syntax-driven Approach for Semantic Role Labeling Yuanhe Tian♥, Han Qin♠, Fei Xia♥, Yan Song♠† ♥University of Washington ♠The Chinese University of Hong Kong (Shenzhen) ♥{yhtian, fxia}@uw.edu ♠hanqin@link.cuhk.edu.cn ♠songyan@cuhk.edu.cn Abstract As an important task to analyze the semantic structure of a sentence, semantic role labeling (SRL) aims to locate the semantic role (e.g., agent) of noun phrases with respect to a given predicate and thus plays an important role in downstream tasks such as dialogue systems. To achieve a better performance in SRL, a model is always required to have a good understanding of the context information. Although one can use an advanced text encoder (e.g., BERT) to capture the context information, extra resources are also required to further improve the model performance. Considering that there are correlations between the syntactic structure and the semantic structure of the sentence, many previous studies leverage auto-generated syntactic knowledge, especially the dependencies, to enhance the modeling of context information through graph-based architectures, where limited attention is paid to other types of auto-generated knowledge. In this paper, we propose map memories to enhance SRL by encoding different types of auto-generated syntactic knowledge (i.e., POS tags, syntactic constituencies, and word dependencies) obtained from off-the-shelf toolkits. Experimental results on two English benchmark datasets for span-style SRL (i.e., CoNLL-2005 and CoNLL-2012) demonstrate the effectiveness of our approach, which outperforms strong baselines and achieves state-of-the-art results on CoNLL-2005.‡ Keywords: semantic role labeling, memory networks, syntactic information 1. Introduction Semantic role labeling (SRL) is an important and fun- damental task in natural language processing (NLP), since it provides a basic analysis of the semantic struc- ture of the input sentence and plays an important role in downstream NLP tasks such as natural language infer- ence (Saha et al., 2020), machine translation (Marcheg- giani et al., 2018), and dialog generation (Xu et al., 2020). To achieve a good SRL performance, it is nor- mally required to have precise understanding of the contextual information for the running text. Therefore, recent studies (T¨ackstr¨om et al., 2015; FitzGerald et al., 2015; Zhou and Xu, 2015; He et al., 2017; Tan et al., 2018; He et al., 2018; Ouchi et al., 2018; Shi and Lin, 2019; Guan et al., 2019; Li et al., 2020a; Li et al., 2020c) applied advanced text encoders (such as Bi- LSTM, Transformer (Vaswani et al., 2017), and pre- trained language models) that are proficient in effec- tively capturing contextual features and achieved great success on this task. Although such encoders are effec- tive, it is always demanding by semantic tasks with ex- tra knowledge. Therefore, to further enhance SRL per- formance, auto-generated knowledge especially syn- tactic ones, has been widely used and demonstrated to be more effective than only using the aforementioned encoders (Lewis et al., 2015; Wang et al., 2019; Kasai et al., 2019; Marcheggiani and Titov, 2020; Fei et al., …

Content truncated to 500 words