Two-View Label Propagation to Semi-supervised Reader Emotion Classification

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 2647–2655, Osaka, Japan, December 11-17 2016. Two-View Label Propagation to Semi-supervised Reader Emotion Classifi- cation Shoushan Li, Jian Xu, Dong Zhang, Guodong Zhou Natural Language Processing Lab School of Computer Science and Technology, Soochow University, China lishoushan@suda.edu.cn, jxu1017@stu.suda.edu.cn dzhang@stu.suda.edu.cn, gdzhou@suda.edu.cn Abstract In the literature, various supervised learning approaches have been adopted to address the task of reader emotion classification. However, the classification performance greatly suffers when the size of the la- beled data is limited. In this paper, we propose a two-view label propagation approach to semi-supervised reader emotion classification by exploiting two views, namely source text and response text in a label propagation algorithm. Specifically, our approach depends on two word-document bipartite graphs to model the relationship among the samples in the two views respectively. Besides, the two bipartite graphs are integrated by linking each source text sample with its corresponding response text sample via a length- sensitive transition probability. In this way, our two-view label propagation approach to semi-supervised reader emotion classification largely alleviates the reliance on the strong sufficiency and independence assumptions of the two views, as required in co-training. Empirical evaluation demonstrates the effec- tiveness of our two-view label propagation approach to semi-supervised reader emotion classification. 1 Introduction Source Text (News): An earthquake of 7.0 magnitude struck China in Lushan County of Ya’an, Sichuan Province, causing seri- ous casualties and property losses, and millions of people in distress…… Writer Emotion: Neutral Reader Emotion: (Sadness), (Worry) Response Text (Comments): (1) Be sure to cherish the golden rescue time. (2) Why there is always an earthquake, so sad, wish all the best. (3) I experienced this quake, all too suddenly, and I will never forget. Figure 1: An example of a news article, together with its writer and reader emotions Emotion classification aims to predict the involving emotion towards a piece of text. For a particular text, there always exist two kinds of emotions, namely writer emotion and reader emotion, where the former concerns the emotion produced by the writer who writes the text and the latter concerns the This work is licensed under a Creative Commons Attribution 4.0 International Licence. Licence details: http://creativecom- mons.org/licenses/by/4.0/  Corresponding author 2647 emotion produced by the reader who reads the text. For instance, in Figure 1, given the news text about earthquake, the writer emotion is more likely to be neutral due to the professionalism of the news re- porter, while the reader emotion might be sadness, or worry. Recent years have seen growing interest in reader emotion classification due to its importance in more and more real-life applications, such as con- tent recommendation and online advertisement. Conventional approaches to reader emotion classification conceptualize the task as a supervised learn- ing problem and rely on a large-scale human-annotated data for model learning. Although such super- vised approaches deliver reasonably good performance, the reliance on labeled data, which is normally difficult and highly expensive to obtain, presents a …

Content truncated to 500 words