A Question Answering Approach for Emotion Cause Extraction

Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 1593–1602 Copenhagen, Denmark, September 7–11, 2017. c⃝2017 Association for Computational Linguistics A Question Answering Approach to Emotion Cause Extraction Lin Guia,b, Jiannan Hua, Yulan Hec, Ruifeng Xua,d†, Qin Lue, Jiachen Dua aShenzhen Graduate School, Harbin Institute of Technology, China bCollege of Mathematics and Computer Science, Fuzhou University, China cSchool of Engineering and Applied Science, Aston University, United Kingdom dGuangdong Provincial Engineering Technology Research Center for Data Science, China eDepartment of Computing, the Hong Kong Polytechnic University, Hong Kong guilin.nlp@gmail.com, hujiannan0526@gmail.com, y.he9@aston.ac.uk, xuruifeng@hit.edu.cn, csluqin@comp.polyu.edu.hk, dujiachen@stmail.hitsz.edu.cn Abstract Emotion cause extraction aims to identify the reasons behind a certain emotion ex- pressed in text. It is a much more diffi- cult task compared to emotion classifica- tion. Inspired by recent advances in using deep memory networks for question an- swering (QA), we propose a new approach which considers emotion cause identifica- tion as a reading comprehension task in QA. Inspired by convolutional neural net- works, we propose a new mechanism to store relevant context in different memory slots to model context information. Our proposed approach can extract both word level sequence features and lexical fea- tures. Performance evaluation shows that our method achieves the state-of-the-art performance on a recently released emo- tion cause dataset, outperforming a num- ber of competitive baselines by at least 3.01% in F-measure. 1 Introduction With the rapid growth of social network platforms, more and more people tend to share their expe- riences and emotions online. Emotion analysis of online text becomes a new challenge in Natu- ral Language Processing (NLP). In recent years, studies in emotion analysis largely focus on emo- tion classification including detection of writers’ emotions (Gao et al., 2013) as well as readers’ emotions (Chang et al., 2015). There are also some information extraction tasks defined in emo- tion analysis (Chen et al., 2016; Balahur et al., 2011), such as extracting the feeler of an emotion (Das and Bandyopadhyay, 2010). These methods †Corresponding Author: xuruifeng@hit.edu.cn assume that emotion expressions are already ob- served. Sometimes, however, we care more about the stimuli, or the cause of an emotion. For in- stance, Samsung wants to know why people love or hate Note 7 rather than the distribution of dif- ferent emotions. Ex.1 我的手机昨天丢了，我现在很难过。 Ex.1 Because I lost my phone yesterday, I feel sad now. In an example shown above, “sad” is an emotion word, and the cause of “sad” is “I lost my phone”. The emotion cause extraction task aims to iden- tify the reason behind an emotion expression. It is a more difficult task compared to emotion classi- fication since it requires a deep understanding of the text that conveys an emotions. Existing approaches to emotion cause extrac- tion mostly rely on methods typically used in information extraction, such as rule based tem- plate matching, sequence labeling and classifica- tion based methods. Most of them use linguistic rules or lexicon features, but do not consider the semantic information and ignore the relation be- …

Content truncated to 500 words