Description of the NTOU Complex QA System

We participated in NTCIR-8 CCLQA Task for Chinese-to-Chinese (C-C) and English-to-Chinese (E-C) tasks. This is our first time developing a QA system for complex questions, although we already have some experiences on handling BIOGRAPHY, DEFINITION, and WHY questions. This paper provides the description of our complex QA system, the NTOU XQA System participated in NTCIR-8 CCLQA Task. This QA system can answer several types of factoid questions and 5 types of complex questions defined in NTCIR-8 CCLQA Task. Different strategies are designed for finding answers to different types of questions. Named entity recognition, distance scores of question keywords, answer information patterns, and search results from the Web are techniques integrated in these strategies. The best F-measure score achieved by our system is 19.88% in monolingual task and 13.62% in cross-lingual task. But unfortunately the official evaluation is incorrect. This paper is structured as follows. Section 2 gives a description of the architecture of our complex QA system, NTOU XQA System. Section 3 gives details of methods to do answer type classification. Section 4 introduces the procedures to find answers from the Internet for factoid and WHY questions. Section 5 expresses how we extract final answers from the ACLIA2 Corpus. Section 6 presents the strategies used in our formal runs and their evaluation results. Section 7 concludes this paper. Category and Subject Descriptions H3.4 [Information Storage and Retrieval] Systems and Software Question-answering (fact retrieval) systems General Terms Performance, Experimentation