Multi-Granularity Semantic Aware Graph Model for Reducing Position Bias in Emotion Cause Pair Extraction

Findings of the Association for Computational Linguistics: ACL 2022, pages 1203 - 1213 May 22-27, 2022 c⃝2022 Association for Computational Linguistics Multi-Granularity Semantic Aware Graph Model for Reducing Position Bias in Emotion-Cause Pair Extraction Yinan Bao1,2, Qianwen Ma1,2, Lingwei Wei1,2, Wei Zhou1,∗and Songlin Hu1,2 1Institute of Information Engineering, Chinese Academy of Sciences 2School of Cyber Security, University of Chinese Academy of Sciences {baoyinan, maqianwen, weilingwei, zhouwei, husonglin}@iie.ac.cn Abstract The Emotion-Cause Pair Extraction (ECPE) task aims to extract emotions and causes as pairs from documents. We observe that the relative distance distribution of emotions and causes is extremely imbalanced in the typi- cal ECPE dataset. Existing methods have set a fixed size window to capture relations be- tween neighboring clauses. However, they ne- glect the effective semantic connections be- tween distant clauses, leading to poor gener- alization ability towards position-insensitive data. To alleviate the problem, we propose a novel Multi-Granularity Semantic Aware Graph model (MGSAG) to incorporate fine- grained and coarse-grained semantic features jointly, without regard to distance limitation. In particular, we first explore semantic de- pendencies between clauses and keywords ex- tracted from the document that convey fine- grained semantic features, obtaining keywords enhanced clause representations. Besides, a clause graph is also established to model coarse- grained semantic relations between clauses. Experimental results indicate that MGSAG surpasses the existing state-of-the-art ECPE models. Especially, MGSAG outperforms other models significantly in the condition of position-insensitive data. 1 Introduction Emotion Cause Analysis (ECA) has attracted in- creasing research interest in recent years (Wei et al., 2020; Sun et al., 2021; Singh et al., 2021; Yu et al., 2021), because of the great potential of applying in consumer review mining, public opinion moni- toring, and online empathetic chatbot building. Its goal is to detect causes or stimuli for a certain emo- tion expressed in text. Emotion Cause Pair Extraction (ECPE) (Xia and Ding, 2019) is a new task related to ECA, which is concerned with causal relationships be- ∗ Corresponding author. Figure 1: The distribution of the relative distance of an emotion clause and a cause clause that comprise a pair in the ECPE dataset (Xia and Ding, 2019). Dist0, Dist1, and Dist2 mean the relative distances between the two clauses are 0, 1, and 2 respectively. Dist > 2 means the relative distances are larger than 2. tween emotions and causes. It’s a much more chal- lenging task. Because we need a comprehensive understanding of document content and structure to perform emotion-cause co-extraction and discrimi- nate emotion-cause clause pairs from negative ones (Wei et al., 2020). As shown in the following ex- ample, an emotion clause c7 and a cause clause c2 construct an emotion-cause pair (c7, c2) which is needed to be extracted by an ECPE model. Example. When the driver was about to start the bus to leave the station (c1), an old lady ran to the front of the bus with a fast speed and sat down on the ground (c2). Passen- gers standing in the front of the bus …

Content truncated to 500 words