Relational Graph Attention Network for Aspect-based Sentiment Analysis

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 3229–3238 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 3229 Relational Graph Attention Network for Aspect-based Sentiment Analysis Kai Wang1, Weizhou Shen1, Yunyi Yang1, Xiaojun Quan1∗, Rui Wang2 1School of Data and Computer Science, Sun Yat-sen University, China 2Alibaba Group, China {wangk73, shenwzh3, yangyy37}@mail2.sysu.edu.cn quanxj3@mail.sysu.edu.cn, masi.wr@alibaba-inc.com Abstract Aspect-based sentiment analysis aims to de- termine the sentiment polarity towards a spe- cific aspect in online reviews. Most recent efforts adopt attention-based neural network models to implicitly connect aspects with opin- ion words. However, due to the complexity of language and the existence of multiple as- pects in a single sentence, these models often confuse the connections. In this paper, we ad- dress this problem by means of effective en- coding of syntax information. Firstly, we de- fine a unified aspect-oriented dependency tree structure rooted at a target aspect by reshaping and pruning an ordinary dependency parse tree. Then, we propose a relational graph attention network (R-GAT) to encode the new tree struc- ture for sentiment prediction. Extensive experi- ments are conducted on the SemEval 2014 and Twitter datasets, and the experimental results confirm that the connections between aspects and opinion words can be better established with our approach, and the performance of the graph attention network (GAT) is significantly improved as a consequence. 1 Introduction Aspect-based sentiment analysis (ABSA) aims at fine-grained sentiment analysis of online af- fective texts such as product reviews. Specifi- cally, its objective is to determine the sentiment polarities towards one or more aspects appear- ing in a single sentence. An example of this task is, given a review great food but the service was dreadful, to determine the polarities towards the aspects food and service. Since the two aspects express quite opposite sen- timents, just assigning a sentence-level sentiment polarity is inappropriate. In this regard, ABSA can provide better insights into user reviews compared with sentence-level sentiment analysis. ∗Corresponding author. Intuitively, connecting aspects with their respec- tive opinion words lies at the heart of this task. Most recent efforts (Wang et al., 2016b; Li et al., 2017; Ma et al., 2017; Fan et al., 2018) resort to assorted attention mechanisms to achieve this goal and have reported appealing results. How- ever, due to the complexity of language mor- phology and syntax, these mechanisms fail occa- sionally. We illustrate this problem with a real review So delicious was the noodles but terrible vegetables, in which the opinion word terrible is closer to the aspect noodles than delicious, and there could be terrible noodles appearing in some other reviews which makes these two words closely asso- ciated. Therefore, the attention mechanisms could attend to terrible with a high weight when eval- uating the aspect noodles. Some other efforts explicitly leverage the syntac- tic structure of a sentence to establish the connec- tions. Among them, early attempts rely on hand- crafted syntactic rules (Qiu et al., 2011; Liu et al., 2013), though they are …

Content truncated to 500 words