The Use of Dependency Relation Graph to Enhance the Term Weighting in Question Retrieval

Proceedings of COLING 2012: Technical Papers, pages 3105–3120, COLING 2012, Mumbai, December 2012. The Use of Dependency Relation Graph to Enhance the Term Weighting in Question Retrieval Weinan Zhang1∗ Zhao-yan Ming2† Yu Zhang1 Liqiang Nie2 Ting Liu1 Tat-Seng Chua2 (1) School of Computer Science and Technology, Harbin Institute of Technology, Harbin, China (2) School of Computing, National University of Singapore, Singapore {wnzhang,zhangyu,tliu}@ir.hit.edu.cn {mingzhaoyan,nieliq,dcscts}@nus.edu.sg ABSTRACT With the emergence of community-based question answering (cQA) services, question retrieval has become an integral part of information and knowledge acquisition. Though existing information retrieval (IR) technologies have been found to be successful for document retrieval, they are less effective for question retrieval due to the inherent characteristics of questions, which have shorter texts. One of the major common drawbacks for the term weighting- based question retrieval models is that they overlook the relations between term pairs when computing their weights. To tackle this problem, we propose a novel term weighting scheme by incorporating the dependency relation cues between term pairs. Given a question, we first construct a dependency graph and compute the relation strength between each term pairs. Next, based on the dependency relation scores, we refine the initial term weights estimated by conventional term weighting approaches. We demonstrate that the proposed term weighting scheme can be seamlessly integrated with popular question retrieval models. Comprehensive experiments well validate our proposed scheme and show that it achieves promising performance as compared to the state-of-the-art methods. TITLE AND ABSTRACT IN ANOTHER LANGUAGE (CHINESE) | | |^ ^ ^• • •• • •é é é{ { {' ' 'X X Xã ã ãU U U? ? ?¯ ¯ ¯é é éu u u¢ ¢ ¢¥ ¥ ¥� � �c c c‘ ‘ ‘D D D� � � ‘X�«.¯‰ÑÖ�Ñy§¯éu¢¤•�&E9•£¼��-‡å»"¦+® k�&Eu¢�.3©�u¢•¡��¤õ§�´du¯éu¢�á©�A5§¦�®k u¢�.éJ·^u¯éu¢" éu®k�Äuc‘D��¯éu¢�.�󧘇̇ �¯K´3OŽc‘�-ž�Ñ�c‘ƒm�'X"•�)ûù‡¯K§·‚JÑ�˜« #�|^c‘m�••é{'XŠ•‚¢�c‘D�Å›"éu‰½¯é§·‚Äk�ï ••'Xã5OŽz‡c‘é�'érݧ?�·‚Šâ••'éÝ5•#~5�c‘� -"·‚�y�¤JÑ�c‘D�Å›U�k�/�Ü�yk�¯éu¢�.¥§…¢� (Jƒ'u�c•`¯éu¢�.k��ŒJ," KEYWORDS: cQA, Question Retrieval, Dependency Relations, Term Weighting. KEYWORDS IN CHINESE: �«.¯‰§¯éu¢§••é{'X§c‘D�. ∗ This work was done when the first author was an intern in National University of Singapore. † Corresponding author 3105 1 Introduction With the proliferation and growth of Web 2.0, cQA services, such as Yahoo! Answers1, Quora2 and WikiAnswer3, have emerged as extremely popular alternatives to acquire information online. They permit information seekers to post their specific questions on any topic and obtain answers provided by other participants. Meanwhile, the blooming social networking technologies quickly link the questions to the experts with first hand experiences and propagate well-answered questions among public who also have similar or relevant questions. Over times, a tremendous number of high quality QA pairs devoted by human intelligence has been accumulated as comprehensive knowledge bases, which greatly facilitate general users to seek information by querying in natural languages (Park and Croft, 2010; Ming et al., 2010; Park et al., 2011). As cQA services contain large scale question and answer (Q&A) archives, they offer an invaluable information resource on the Web, to provide answers to new questions posed by the users (Jeon et al., 2005b). However, question retrieval is not a trivial task (Wang et al., …

Content truncated to 500 words