A Multi-turn Machine Reading Comprehension Framework with Rethink Mechanism for Emotion-Cause Pair Extraction

Proceedings of the 29th International Conference on Computational Linguistics, pages 6726–6735 October 12–17, 2022. 6726 A Multi-turn Machine Reading Comprehension Framework with Rethink Mechanism for Emotion-Cause Pair Extraction Changzhi Zhou, Dandan Song∗, Jing Xu, Zhijing Wu School of Computer Science and Technology, Southeast Academy of Information Technology, Beijing Institute of Technology, Beijing, China zhou_changzhi97@163.com sdd@bit.edu.cn, xujing@bit.edu.cn wuzhijing.joyce@gmail.com Abstract Emotion-cause pair extraction (ECPE) is an emerging task in emotion cause analysis, which extracts potential emotion-cause pairs from an emotional document. Most recent studies use end-to-end methods to tackle the ECPE task. However, these methods either suffer from a label sparsity problem or fail to model compli- cated relations between emotions and causes. Furthermore, they all do not consider explicit semantic information of clauses. To this end, we transform the ECPE task into a document- level machine reading comprehension (MRC) task and propose a Multi-turn MRC framework with Rethink mechanism (MM-R). Our frame- work can model complicated relations between emotions and causes while avoiding generating the pairing matrix (the leading cause of the la- bel sparsity problem). Besides, the multi-turn structure can fuse explicit semantic information flow between emotions and causes. Extensive experiments on the benchmark emotion cause corpus demonstrate the effectiveness of our pro- posed framework, which outperforms existing state-of-the-art methods.1 1 Introduction Emotion cause extraction (ECE) is a classical emo- tion cause analysis task that aims to extract the corresponding causes of the given emotional ex- pressions in an emotional document (Gui et al., 2016). However, the ECE task is not practical in real-world scenarios without annotated emotions. To overcome the limitation and capture mutual indications of emotions and causes together, Xia and Ding (2019) came up with a new task called emotion-cause pair extraction (ECPE), which aims to extract a potential emotion-cause pair set consist- ing of all emotions and their corresponding causes from a document. As shown in Figure 1(a), c2 is ∗Corresponding Author 1Data and code are available at https://github.com/zhoucz97/ECPE-MM-R c1: Mike said, c2: I am so happy (emotion), c3: because I have my own family now (cause), c4: I thank all the people who have helped me (emotion, cause). Input: a document Output: an emotion-cause pair set { (c2, c3), (c4, c4) } (a) An example of the ECPE task (c1, c1) (c1, c2) (c1, c3) (c1, c4) (c2, c1) (c2, c2) (c2, c3) (c2, c4) (c3, c1) (c3, c2) (c3, c3) (c3, c3) (c4, c1) (c4, c2) (c4, c3) (c4, c4) (b) Pairing matrix c2, c4 c1, c2, c3, c4 (c2, c3), (c2, c4), (c4, c4) (c2, c3), (c4, c4) (c) Our approach Figure 1: The green colour denotes an emotion clause, and the red colour denotes a cause clause. Figure (a) is an example of the ECPE task. Figure (b) is a pairing matrix generated by pair-level end-to-end approaches. Only (c2, c3) and (c4, c4) are valid pairs. Figure (c) shows the processing results by our MM-R in each turn. an emotion clause, c3 is the corresponding cause clause, and c4 is an emotion …

Content truncated to 500 words