Boundary Detection with BERT for Span-level Emotion Cause Analysis

Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 676–682 August 1–6, 2021. ©2021 Association for Computational Linguistics 676 Boundary Detection with BERT for Span-level Emotion Cause Analysis Xiangju Li1, Wei Gao2, Shi Feng1, Yifei Zhang1, Daling Wang1 1School of Computer Science and Engineering, Northeastern University, Shenyang, China 2School of Computing and Information Systems, Singapore Management University, Singapore lixiangju100@163.com; weigao@smu.edu.sg; fengshi@cse.neu.edu.cn; zhangyifei@cse.neu.edu.cn; wangdaling@cse.neu.edu.cn Abstract Emotion cause analysis (ECA) has been an emerging topic in natural language processing, which aims to identify the reasons behind a certain emotion expressed in the text. Most ECA methods intend to identify the clause which contains the cause of a given emotion, but such clause-level ECA (CECA) can be am- biguous and imprecise. In this paper, we aim at span-level ECA (SECA) by detecting the precise boundaries of text spans conveying ac- curate emotion causes from the given context. We formulate this task as sequence labeling and position identification problems and de- sign two neural methods to solve them. Exper- iments on two benchmark ECA datasets show that the proposed methods substantially outper- form the existing ECA models 1. 1 Introduction The task of emotion cause analysis (ECA) (Lee et al., 2010a), which is to extract the causes of an emotion expression from a given context, has gained increasing attention recently. Most existing studies (Gui et al., 2016, 2017; Li et al., 2018, 2019; Xia et al., 2019) formulate ECA as a clause-level task (dubbed as CECA), which tries to extract the clauses that contain the emotion cause content. In CECA, a clause is typically a text segment separated by punctuation marks (e.g., ‘,’, ‘.’, ‘?’, ‘!’, etc.) in the given context. In the following example, the clause [x3] will be extracted as it contains the reason “the risk of infringement” that stimulates the emotion afraid. Example 1. 2 [x1] The claim for damages caused by infringement has legal basis, [x2] and it is logi- cal and reasonable. [x3] We can’t give up the good 1The code and datasets can be found at https://github.com/xxxyyy2020/boundary-master 2Bold: Gold emotion cause span; Underline: emotion expression. [xi] represents the i-th clause. The original report is in Chinese and translated into English. character of friendship and mutual assistance just because we are afraid of the risk of infringement. [x4] Give up food for being afraid of being choked. [x5] We should adhere to the traffic rules. However, determining the clause containing the stimulus is sub-optimal and inaccurate for ECA. In Example 1, while among the 5 clauses [x3] is the best, its main content “We can’t give up .. .assistance” is not the cause of afraid. Such gap motivates a strong need of pinpointing more pre- cise or finer-grained cause expressions which can convey the specific reasons of an emotion. Some studies (Gao et al., 2015a,b; Neviarouskaya and Aono, 2013) try to ex- tract emotion cause triples (a triple like (noun, verb, noun)) or emotion cause phrases. The extracted triples or phrases are usually not a complete emotion cause …

Content truncated to 500 words