Learning Span-Level Interactions for Aspect Sentiment Triplet Extraction

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 4755–4766 August 1–6, 2021. ©2021 Association for Computational Linguistics 4755 Learning Span-Level Interactions for Aspect Sentiment Triplet Extraction Lu Xu* 1, 2, Yew Ken Chia* 1, 2, Lidong Bing2 1 Singapore University of Technology and Design 2 DAMO Academy, Alibaba Group {xu lu,yewken chia}@mymail.sutd.edu.sg l.bing@alibaba-inc.com Abstract Aspect Sentiment Triplet Extraction (ASTE) is the most recent subtask of ABSA which outputs triplets of an aspect target, its asso- ciated sentiment, and the corresponding opin- ion term. Recent models perform the triplet extraction in an end-to-end manner but heav- ily rely on the interactions between each tar- get word and opinion word. Thereby, they cannot perform well on targets and opinions which contain multiple words. Our proposed span-level approach explicitly considers the interaction between the whole spans of tar- gets and opinions when predicting their sen- timent relation. Thus, it can make predic- tions with the semantics of whole spans, en- suring better sentiment consistency. To ease the high computational cost caused by span enumeration, we propose a dual-channel span pruning strategy by incorporating supervision from the Aspect Term Extraction (ATE) and Opinion Term Extraction (OTE) tasks. This strategy not only improves computational effi- ciency but also distinguishes the opinion and target spans more properly. Our framework si- multaneously achieves strong performance for the ASTE as well as ATE and OTE tasks. In particular, our analysis shows that our span- level approach achieves more significant im- provements over the baselines on triplets with multi-word targets or opinions. 1 1 Introduction Aspect-Based Sentiment Analysis (ABSA) (Liu, 2012; Pontiki et al., 2014) is an aggregation of sev- eral fine-grained sentiment analysis tasks, and its various subtasks are designed with the aspect tar- get as the fundamental item. For the example in ∗ Equal contribution. Lu Xu and Yew Ken Chia are under the Joint PhD Program between Alibaba and Singapore University of Technology and Design. 1We make our code publicly available at https:// github.com/chiayewken/Span-ASTE. - - Figure 1: An example of ASTE. The spans highlighted in orange are target terms, and the span in blue is opin- ion term. The “-” on top of target terms indicates nega- tive sentiment. Figure 1, the aspect targets are “Windows 8” and “touchscreen functions”. Aspect Sentiment Classi- fication (ASC) (Dong et al., 2014; Zhang et al., 2016; Yang et al., 2017; Li et al., 2018a; Tang et al., 2019) is one of the most well-explored subtasks of ABSA and aims to predict the sentiment polarity of a given aspect target. However, it is not always practical to assume that the aspect target is pro- vided. Aspect Term Extraction (ATE) (Yin et al., 2016; Li et al., 2018b; Ma et al., 2019) focuses on extracting aspect targets, while Opinion Term Extraction (OTE) (Yang and Cardie, 2012; Klinger and Cimiano, 2013; Yang and Cardie, 2013) aims to extract the opinion terms which largely determine the sentiment …

Content truncated to 500 words