Seeking Common but Distinguishing Difference, A Joint Aspect-based Sentiment Analysis Model

Seeking Common but Distinguishing Difference, A Joint Aspect-based Sentiment Analysis Model Hongjiang Jing1,2,3†, Zuchao Li1,2,3, Hai Zhao1,2,3∗, and Shu Jiang1,2,3† 1Department of Computer Science and Engineering, Shanghai Jiao Tong University 2Key Laboratory of Shanghai Education Commission for Intelligent Interaction and Cognitive Engineering, Shanghai Jiao Tong University, Shanghai, China 3MoE Key Lab of Artificial Intelligence, AI Institute, Shanghai Jiao Tong University {jinghj,charlee,jshmjs45}@sjtu.edu.cn, zhaohai@cs.sjtu.edu.cn Abstract Aspect-based sentiment analysis (ABSA) task consists of three typical subtasks: aspect term extraction, opinion term extraction, and senti- ment polarity classification. These three sub- tasks are usually performed jointly to save re- sources and reduce the error propagation in the pipeline. However, most of the existing joint models only focus on the benefits of encoder sharing between subtasks but ignore the dif- ference. Therefore, we propose a joint ABSA model, which not only enjoys the benefits of encoder sharing but also focuses on the differ- ence to improve the effectiveness of the model. In detail, we introduce a dual-encoder design, in which a pair encoder especially focuses on candidate aspect-opinion pair classification, and the original encoder keeps attention on se- quence labeling. Empirical results show that our proposed model shows robustness and sig- nificantly outperforms the previous state-of- the-art on four benchmark datasets. 1 Introduction Sentiment analysis is a task that aims to retrieve the sentiment polarity based on three levels of gran- ularities: document level, sentence level, and en- tity and aspect level (Liu, 2012), which is under the urgent demands of several society scenarios (Preethi et al., 2017; Cobos et al., 2019; Islam and Zibran, 2017; Novielli et al., 2018). Recently, the aspect-based sentiment analysis (ABSA) task (Pon- tiki et al., 2014), focusing on excavating the spe- cific aspect from an annotated review, has aroused much attention from researchers, in which this pa- per mainly concerns the aspect/opinion term extrac- tion and sentiment classification task. The latest benchmark proposed by Peng et al. (2020) formu- lates the relevant information into a triplet: target aspect object, opinion clue, and sentiment polarity ∗Corresponding author. †Equal contribution. ������������ ��������������� ������������ ���������������� ����� ���������� ������������������������������������� ���������������������� ����������������������������� �������������� ��������������� ������������������������������������������ �������� ���������������������������������� ������������������������������������������������������ ��������������� ���������������������� ������������ The view is spectacular, and the food is great. positive positive ���������������� ����������������������� ����������������������� ���������������� ����������������� ����������������� ��������� Figure 1: The subtasks in our proposed model. orientation. Thus, the concerned aspect term ex- traction becomes a task of Aspect Sentiment Triplet Extraction (ASTE). Similarly, the relevant informa- tion is formulated into a pair with aspect term and sentiment polarity, and the task is defined as As- pect Term Extraction and Sentiment Classification (AESC). Figure 1 shows an example of ASTE and AESC. Two early methods handle the triplet extraction task efficiently (Zhang et al., 2020a; Huang et al., 2021). Both are typically composed of a sequence representation layer to predict the aspect/opinion term mentions and a classification layer to infer the sentiment polarity of the predicted mention pair of the last layer. However, as is often the case, such model design may easily …

Content truncated to 500 words