Function-Based Question Classification for General QA

Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 1119–1128, MIT, Massachusetts, USA, 9-11 October 2010. c⃝2010 Association for Computational Linguistics Function-based question classification for general QA Fan Bu, Xingwei Zhu, Yu Hao and Xiaoyan Zhu State Key Laboratory of Intelligent Technology and Systems Tsinghua National Laboratory for Information Science and Technology Department of Computer Sci. and Tech., Tsinghua University buf08@mails.tsinghua.edu.cn etzhu192@hotmail.com haoyu@mail.tsinghua.edu.cn zxy-dcs@tsinghua.edu.cn Abstract In contrast with the booming increase of inter- net data, state-of-art QA (question answering) systems, otherwise, concerned data from spe- cific domains or resources such as search en- gine snippets, online forums and Wikipedia in a somewhat isolated way. Users may welcome a more general QA system for its capability to answer questions of various sources, inte- grated from existed specialized sub-QA en- gines. In this framework, question classifica- tion is the primary task. However, the current paradigms of question classification were focused on some speci- fied type of questions, i.e. factoid questions, which are inappropriate for the general QA. In this paper, we propose a new question clas- sification paradigm, which includes a ques- tion taxonomy suitable to the general QA and a question classifier based on MLN (Markov logic network), where rule-based methods and statistical methods are unified into a single framework in a fuzzy discriminative learning approach. Experiments show that our method outperforms traditional question classification approaches. 1 Introduction During a long period of time, researches on question answering are mainly focused on finding short and concise answers from plain text for factoid questions driven by annual trackes such as CLEF, TREC and NTCIR. However, people usually ask more complex questions in real world which cannot be handled by these QA systems tailored to factoid questions. During recent years, social collaborative applica- tions begin to flourish, such as Wikipedia, Facebook, Yahoo! Answers and etc. A large amount of semi- structured data, which has been accumulated from these services, becomes new sources for question answering. Previous researches show that different sources are suitable for answering different ques- tions. For example, the answers for factoid questions can be extracted from webpages with high accuracy, definition questions can be answered by correspond- ing articles in wikipedia(Ye et al., 2009) while com- munity question answering services provide com- prehensive answers for complex questions(Jeon et al., 2005). It will greatly enhance the overall per- formance if we can classify questions into several types, distribute each type of questions to suitable sources and trigger corresponding strategy to sum- marize returned answers. Question classification (QC) in factoid QA is to provide constraints on answer types that allows fur- ther processing to pinpoint and verify the answer (Li and Roth, 2004). Usually, questions are classi- fied into a fine grained content-based taxonomy(e.g. UIUC taxonomy (Li and Roth, 2002)). We can- not use these taxonomies directly. To guide ques- tion distribution and answer summarization, ques- tions are classified according to their functions in- stead of contents. Motivated by related work on user goal classi- fication(Broder, 2002; Rose …

Content truncated to 500 words