Learning the Latent Topics for Question Retrieval in Community QA

Proceedings of the 5th International Joint Conference on Natural Language Processing, pages 273–281, Chiang Mai, Thailand, November 8 – 13, 2011. c⃝2011 AFNLP Learning the Latent Topics for Question Retrieval in Community QA Li Cai, Guangyou Zhou, Kang Liu, and Jun Zhao National Laboratory of Pattern Recognition Institute of Automation, Chinese Academy of Sciences 95 Zhongguancun East Road, Beijing 100190, China {lcai,gyzhou,kliu,jzhao}@nlpr.ia.ac.cn Abstract Community-based Question Answering (cQA) is a popular online service where users can ask and answer questions on any topics. This paper is concerned with the problem of question retrieval. Ques- tion retrieval in cQA aims to find histori- cal questions that are semantically equiva- lent or relevant to the queried questions. Although the translation-based language model (Xue et al., 2008) has gained the state-of-the-art performance for question retrieval, they ignore the latent topic in- formation in calculating the semantic sim- ilarity between questions. In this paper, we propose a topic model incorporated with the category information into the pro- cess of discovering the latent topics in the content of questions. Then we com- bine the semantic similarity based latent topics with the translation-based language model into a unified framework for ques- tion retrieval. Experiments are carried out on a real world cQA data set from Ya- hoo! Answers. The results show that our proposed method can significantly im- prove the question retrieval performance of translation-based language model. 1 Introduction Over the past few years, large scale question and answer archives have become an important infor- mation resource on the Web. These include the tra- ditional FAQ archives constructed by the experts or companies for their products and the emerging community-based online services, such as Yahoo! Answers1 and Live QnA2. The major challenge for cQA retrieval is the lex- ical gap (or lexical chasm) between the queried 1http://answers.yahoo.com 2http://qna.live.com questions and the question-answer pairs in the archives (Jeon et al., 2005; Xue et al., 2008). To solve the lexical gap problem, most researchers regarded the question retrieval task as a statisti- cal machine translation problem by using IBM model 1 (Brown et al., 1993) to learn the word- to-word translation probabilities (Berger and Laf- ferty, 1999; Jeon et al., 2005; Xue et al., 2008; Lee et al., 2008; Bernhard and Gurevych, 2009; Cao et al., 2010). Although the translation-based lan- guage model(TRLM) has yielded the state-of-the- art performance for question retrieval, they model the word translation probabilities without taking into account the distribution of words in the whole content. In this paper, we argue that it is beneficial to ex- ploit the latent topic information for question re- trieval. The basic idea is as follows: first we em- ploy the topic model (e.g., LDA) to discover the latent topics in the content of questions, and cal- culate the semantic similarity between questions based on the latent topic information. Moreover, a distinctive feature of question-answer archives in cQA is that cQA services always organize ques- tions into a hierarchy of categories. We propose an improved latent topic model by …

Content truncated to 500 words