Modelling Context and Syntactical Features for Aspect-based Sentiment Analysis

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 3211–3220 July 5 - 10, 2020. c⃝2020 Association for Computational Linguistics 3211 Modelling Context and Syntactical Features for Aspect-based Sentiment Analysis Minh Hieu Phan & Philip Ogunbona School of Computer Science and Software Engineering University of Wollongong, NSW 2522, Australia vmhp806@uowmail.edu.au, philipo@uow.edu.au Abstract The aspect-based sentiment analysis (ABSA) consists of two conceptual tasks, namely an aspect extraction and an aspect sentiment clas- sification. Rather than considering the tasks separately, we build an end-to-end ABSA so- lution. Previous works in ABSA tasks did not fully leverage the importance of syntac- tical information. Hence, the aspect extrac- tion model often failed to detect the bound- aries of multi-word aspect terms. On the other hand, the aspect sentiment classifier was un- able to account for the syntactical correlation between aspect terms and the context words. This paper explores the grammatical aspect of the sentence and employs the self-attention mechanism for syntactical learning. We com- bine part-of-speech embeddings, dependency- based embeddings and contextualized embed- dings (e.g. BERT, RoBERTa) to enhance the performance of the aspect extractor. We also propose the syntactic relative distance to de-emphasize the adverse effects of unrelated words, having weak syntactic connection with the aspect terms. This increases the accuracy of the aspect sentiment classifier. Our solu- tions outperform the state-of-the-art models on SemEval-2014 dataset in both two subtasks. 1 Introduction The process of understanding the sentiments ex- pressed by consumers in a product review (opin- ionated text) is referred to as sentiment analysis. Deep insights into the opinionated text are gained through a fine-grained entity- or aspect-based senti- ment labeling of the product being reviewed. Such insights can be invaluable for business decision making. Aspect-based sentiment analysis (ABSA) con- sists of two sub-tasks, namely an aspect extrac- tion (AE) and an aspect sentiment classification (ASC). However, the majority of reported works focused on one of the two sub-tasks alone. Rep- resentative works include (Xu et al., 2018; Da’u and Salim, 2019; Poria et al., 2016) for aspect ex- traction and (Zeng et al., 2019; Huang et al., 2018; Song et al., 2019; Thet et al., 2010) for aspect sen- timent classification. Recent approaches (He et al., 2019; Wang et al., 2018; Li et al., 2019) attempted to develop an integrated solution to solve both tasks simultaneously by formulating both sub-tasks as a single sequence labelling with a unified tagging scheme. Adding unified tokens introduces over- head and complexity in the original ABSA tasks. Thus, multi-task models often have poorer perfor- mance compared with single-task models which are trained independently. Recent advances in the NLU introduced contex- tualized language models, namely OpenAI GPT (Radford et al., 2018), BERT (Devlin et al., 2018) and RoBERTa (Liu et al., 2019). These models can capture the characteristics of word uses and account for different textual context in which words appear. Upon investigating the latest BERT/RoBERTa- based architectures used in aspect extraction, it became apparent that they were unable to deter- mine …

Content truncated to 500 words