Does syntax matter? A strong baseline for Aspect-based Sentiment Analysis with RoBERTa

Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1816–1829 June 6–11, 2021. ©2021 Association for Computational Linguistics 1816 Does syntax matter? A strong baseline for Aspect-based Sentiment Analysis with RoBERTa Junqi Dai1,∗, Hang Yan1,∗, Tianxiang Sun1, Pengfei Liu2, Xipeng Qiu1,† 1Shanghai Key Laboratory of Intelligent Information Processing, Fudan University 1School of Computer Science, Fudan University 2Carnegie Mellon University {jqdai19,hyan19,txsun19,xpqiu}@fudan.edu.cn pliu3@cs.cmu.edu Abstract Aspect-Based Sentiment Analysis (ABSA), aiming at predicting the polarities for aspects, is a fine-grained task in the field of sentiment analysis. Previous work showed syntactic in- formation, e.g. dependency trees, can effec- tively improve the ABSA performance. Re- cently, pre-trained models (PTMs) also have shown their effectiveness on ABSA. There- fore, the question naturally arises whether PTMs contain sufficient syntactic information for ABSA so that we can obtain a good ABSA model only based on PTMs. In this paper, we firstly compare the induced trees from PTMs and the dependency parsing trees on several popular models for the ABSA task, showing that the induced tree from fine- tuned RoBERTa (FT-RoBERTa) outperforms the parser-provided tree. The further analy- sis experiments reveal that the FT-RoBERTa Induced Tree is more sentiment-word-oriented and could benefit the ABSA task. The experi- ments also show that the pure RoBERTa-based model can outperform or approximate to the previous SOTA performances on six datasets across four languages since it implicitly in- corporates the task-oriented syntactic informa- tion.1 1 Introduction Aspect-based sentiment analysis (ABSA) aims to do the fine-grained sentiment analysis towards as- pects (Pontiki et al., 2014, 2016). Specifically, for one or more aspects in a sentence, the task calls for detecting the sentiment polarities for all aspects. Take the sentence “great food but the service was dreadful” for example, the task is to predict the sen- timents towards the underlined aspects, which ex- pects to get polarity positive for aspect food and po- larity negative for aspect service. Generally, ABSA ∗Equal contribution. †Corresponding author. 1Our code will be released at https://github.com/ ROGERDJQ/RoBERTaABSA. contains aspect extraction (AE) and aspect-level sentiment classification (ALSC). We only focus on the ALSC task. Early works of ALSC mainly rely on manu- ally designed syntactic features, which is labor- intensive yet insufficient. In order to avoid de- signing hand-crafted features (Jiang et al., 2011; Kiritchenko et al., 2014), various neural network models have been proposed in ALSC (Dong et al., 2014; Vo and Zhang, 2015; Wang et al., 2016; Chen et al., 2017; He et al., 2018; Zhang et al., 2019b; Wang et al., 2020). Since the dependency tree can help the aspects find their contextual words, most of the recently proposed State-of-the-art (SOTA) ALSC models utilize the dependency tree to as- sist in modeling connections between aspects and their opinion words (Wang et al., 2020; Sun et al., 2019b; Zhang et al., 2019b). Generally, these dependency tree based ALSC models are imple- mented in three methods. The first one is to use the topological structure of the dependency tree …

Content truncated to 500 words