Identifying Intention Posts in Discussion Forums

Proceedings of NAACL-HLT 2013, pages 1041–1050, Atlanta, Georgia, 9–14 June 2013. c⃝2013 Association for Computational Linguistics Identifying Intention Posts in Discussion Forums Zhiyuan Chen, Bing Liu Meichun Hsu, Malu Castellanos, Riddhiman Ghosh Department of Computer Science HP Labs University of Illinois at Chicago Palo Alto, CA 94304, USA Chicago, IL 60607, USA {meichun.hsu, malu.castellanos, riddhiman.ghosh}@hp.com czyuanacm@gmail.com,liub@cs.uic.edu Abstract This paper proposes to study the problem of identifying intention posts in online discus- sion forums. For example, in a discussion fo- rum, a user wrote “I plan to buy a camera,” which indicates a buying intention. This in- tention can be easily exploited by advertisers. To the best of our knowledge, there is still no reported study of this problem. Our research found that this problem is particularly suited to transfer learning because in different do- mains, people express the same intention in similar ways. We then propose a new transfer learning method which, unlike a general transfer learning algorithm, exploits several special characteristics of the problem. Exper- imental results show that the proposed meth- od outperforms several strong baselines, including supervised learning in the target domain and a recent transfer learning meth- od. 1 Introduction Social media content is increasingly regarded as an information gold mine. Researchers have stud- ied many problems in social media, e.g., senti- ment analysis (Pang & Lee, 2008; Liu, 2010) and social network analysis (Easley & Kleinberg, 2010). In this paper, we study a novel problem which is also of great value, namely, intention identification, which aims to identify discussion posts expressing certain user intentions that can be exploited by businesses or other interested parties. For example, one user wrote, “I am looking for a brand new car to replace my old Ford Focus”. Identifying such intention automatically can help social media sites to decide what ads to display so that the ads are more likely to be clicked. This work focuses on identifying user posts with explicit intentions. By explicit we mean that the intention is explicitly stated in the text, no need to deduce (hidden or implicit intention). For example, in the above sentence, the author clearly expressed that he/she wanted to buy a car. On the other hand, an example of an implicit sentence is “Anyone knows the battery life of iPhone?” The person may or may not be thinking about buying an iPhone. To our knowledge, there is no reported study of this problem in the context of text documents. The main related work is in Web search, where user (or query) intent classification is a major issue (Hu et al., 2009; Li, 2010; Li, Wang, & Acero, 2008). Its task is to determine what the user is searching for based on his/her keyword queries (2 to 3 words) and his/her click data. We will discuss this and other related work in Section 2. We formulate the proposed problem as a two- class classification problem since an application may only be interested in a particular intention. We define intention posts (positive class) …

Content truncated to 500 words