Active Sentiment Domain Adaptation

Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1701–1711 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1156 Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1701–1711 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1156 Active Sentiment Domain Adaptation Fangzhao Wu†, Yongfeng Huang†∗, and Jun Yan‡ †Department of Electronic Engineering, Tsinghua University ‡Microsoft Research Asia, Beijing, China wufangzhao@gmail.com, yfhuang@tsinghua.edu.cn, junyan@microsoft.com Abstract Domain adaptation is an important techno- logy to handle domain dependence pro- blem in sentiment analysis field. Existing methods usually rely on sentiment classi- fiers trained in source domains. Howe- ver, their performance may heavily decline if the distributions of sentiment features in source and target domains have signifi- cant difference. In this paper, we propose an active sentiment domain adaptation ap- proach to handle this problem. Instead of the source domain sentiment classifiers, our approach adapts the general-purpose sentiment lexicons to target domain with the help of a small number of labeled samples which are selected and annota- ted in an active learning mode, as well as the domain-specific sentiment similarities among words mined from unlabeled sam- ples of target domain. A unified model is proposed to fuse different types of senti- ment information and train sentiment clas- sifier for target domain. Extensive expe- riments on benchmark datasets show that our approach can train accurate sentiment classifier with less labeled samples. 1 Introduction Sentiment classification is widely known as a domain-dependent problem (Liu, 2012; Pang and Lee, 2008; Blitzer et al., 2007; Pan et al., 2010). This is because different domains usually have many different sentiment expressions. For exam- ple, “lengthy” and “boring” are popularly used in Book domain to express negative sentiment. Ho- wever, they are rare in Kitchen appliance domain. Moreover, the same word or phrase may convey ∗Corresponding author. different sentiments in different domains. For in- stance, “unpredictable” is frequently used to ex- press positive sentiment in Movie domain (e.g., “The plot of this movie is fun and unpredictable”). However, it tends to be used as a negative word in Kitchen appliance domain (e.g., “Even holding heat is unpredictable. It is just terrible!”). Thus, every domain has many domain-specific sentiment expressions, which cannot be captured by other domains. The performance of directly applying a general sentiment classifier or a sentiment classi- fier trained in other domains to target domain is usually suboptimal. Since there are a large number of domains in user-generated content, it is impractical to ma- nually annotate enough samples for each dom- ain to train an accurate domain-specific sentiment classifier. Thus, sentiment domain adaptation, which transfers the sentiment classifier trained in a source domain with sufficient labeled data to a target domain with no or scarce labeled data, has been widely studied (Blitzer et al., 2007; Pan et al., 2010; He et al., 2011; Glorot et al., 2011). Existing sentiment domain adaptation methods are mainly based on transfer learning techniques. Many …

Content truncated to 500 words