Identification of relations between answers with global constraints for Community-based Question Answering services

Proceedings of the 5th International Joint Conference on Natural Language Processing, pages 920–928, Chiang Mai, Thailand, November 8 – 13, 2011. c⃝2011 AFNLP Identification of relations between answers with global constraints for Community-based Question Answering services Hikaru Yokono Precision and Intelligence Laboratory, Tokyo Institute of Technology yokono@lr.pi.titech.ac.jp Takaaki Hasegawa NTT Cyber Space Laboratories, NTT Corporation hasegawa.takaaki@lab.ntt.co.jp Genichiro Kikui∗ NTT Cyber Space Laboratories, NTT Corporation kikui.genichiro@lab.ntt.co.jp Manabu Okumura Precision and Intelligence Laboratory, Tokyo Institute of Technology oku@pi.titech.ac.jp Abstract Community-based Question Answering services contain many threads consisting of a question and its answers. When there are many answers for a question, it is hard for a user to understand them all. To address this problem, we focus on logi- cal relations between answers in a thread and present a model for identifying the relations between the answers. We con- sider that there are constraints among the relations, such as a transitive law, and that it might be useful to take these con- straints into account. To consider these constraints, we propose the model based on a Markov logic network. We also in- troduce super-relations to give additional information for logical relation identifica- tion into our model. Through the experi- ment, we show that global constraints and super-relations make it easier to identify the relations. 1 Introduction Community-based Question Answering services, such as Yahoo! Answers1, OKWave2 and Baidu Zhidao3, have become popular web services. In these services, a user posts a question and other users answer it. The questioner chooses one of the answers as the best answer. These services have many threads consisting of one question and a number of answers, and the number of threads ∗Current affiliation is Faculty of Computer Science and System Engineering, Okayama Prefectural University. 1http://answers.yahoo.com/ 2http://okwave.jp/ 3http://zhidao.baidu.com/ grows day by day. The threads are stored and any- one can read them. When a user has a question, if there is a similar question in the service, he or she can refer to the answers to the similar ques- tion. Herefrom, these services are useful for not only the questioner but also other users having a similar question. When more answers get posted for a question, the answers in the thread might become more di- verse. Some of these answers will be similar or oppositional to each other. Also, when a ques- tion tends to have various answers, e.g. the ques- tioner asks for opinions (e.g. “What are your song recommendations?”), it is insufficient to read only the best answer. Generally, as the only one best answer is chosen from the answers, a user may miss other beneficial answers. When a user checks these services with a mobile device, its small dis- play is inefficient to browse all answers. To alleviate these problems, it would be useful to get an overview of answers in a thread, such as by identifying the relations between the answers or by summarizing them (Jimbo et al., 2010; Liu et al., 2008). The purpose of this study is to iden- tify logical relations …

Content truncated to 500 words