Sentence-level Emotion Classification with Label and Context Dependence

Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing, pages 1045–1053, Beijing, China, July 26-31, 2015. c⃝2015 Association for Computational Linguistics Sentence-level Emotion Classification with Label and Context Dependence Shoushan Li†‡, Lei Huang†, Rong Wang†, Guodong Zhou†* †Natural Language Processing Lab, Soochow University, China ‡ Collaborative Innovation Center of Novel Software Technology and Industrialization {shoushan.li, lei.huang2013, wangrong2022}@gmail.com, gdzhou@suda.edu.cn Abstract Predicting emotion categories, such as anger, joy, and anxiety, expressed by a sentence is challenging due to its inherent multi-label classification difficulty and data sparseness. In this paper, we address above two chal- lenges by incorporating the label dependence among the emotion labels and the context de- pendence among the contextual instances into a factor graph model. Specifically, we recast sentence-level emotion classification as a fac- tor graph inferring problem in which the label and context dependence are modeled as vari- ous factor functions. Empirical evaluation demonstrates the great potential and effective- ness of our proposed approach to sentence- level emotion classification. 1 1 Introduction Predicting emotion categories, such as anger, joy, and anxiety, expressed by a piece of text encom- passes a variety of applications, such as online chatting (Galik et al., 2012), news classification (Liu et al., 2013) and stock marketing (Bollen et al., 2011). Over the past decade, there has been a substantial body of research on emotion classifi- cation, where a considerable amount of work has focused on document-level emotion classification. Recently, the research community has become increasingly aware of the need on sentence-level emotion classification due to its wide potential ap- plications, e.g. the massively growing importance of analyzing short text in social media (Ki- ritchenko et al., 2014; Wen and Wan, 2014). In general, sentence-level emotion classification ex- hibits two challenges. 1 * Corresponding author …… <S1>她们都睡了，我蹑手蹑脚摸黑上了 床，凑上去想亲嫣一下，她突然一个转身， 小手‘啪’地搭在了我的脸颊上。</S1> <S2> 现在我终于如愿以偿。</S2> <S3>感受着小手 的温度，享受着这份她对我的依恋，生怕动 一下，会让她的小手离我而去。</S3>…… (English: …… <S1> The girls fall to sleep, so I make my way noiselessly onto the bed, wishing I could get a chance to give a kiss to Yan, suddenly she turn over to me and her little soft hand fall onto my face.</S1> <S2>Praise the Lord, that is all I want.</S2> <S3>Feeling the warm of her hand and the attachment she hold to me, I couldn’t af- ford to move even a little, fearing I may lost her hand.</S3>)……) ------------------------------------------------------------------- Sentence-level Emotion Classification  Input: S1, S2, S3  Output: S1 : joy, love S2: joy S3: joy, love, anxiety Figure 1: An example of a paragraph and the sentences therein with their emotion categories from the corpus collected by Quan and Ren (2009) On one hand, like document-level emotion classification, sentence-level emotion classifica- tion is naturally a multi-label classification prob- lem. That is, each sentence might involve more than one emotion category. For example, as shown in Figure 1, in one paragraph, two sen- tences, i.e., S1 and S3, have two and three emotion categories respectively. Automatically classifying instances with multiple possible categories is 1045 sometimes …

Content truncated to 500 words