A study on question answering system using integrated retrieval method

In recent years, Question Answering (QA) system has been researched extensively and automatic question answering has become an interesting research field and resulted in a visible improvement in its performance. Especially, during the last decade, a number of automatic QA systems have emerged, which has been largely driven by the TREC (Text REtrieval Conference) QA Track. The technology of QA relates to a lot of aspects of NLP (Natural Language Processing), such as Information Retrieval (IR), Information Extraction (IE), Automatic Summarization, Conversation Interface, etc. However, recently the QA systems have emerged following two directions: one direction is to use the TREC QA data as a testing corpus and develop their own search engines and answer extraction techniques on top of the corpus; Another direction is to use the WWW as the potential answer source and use generic search engines, such as Google, to extract the answers for the users’ questions. Although, the current trend in Question Answering focus on open domain, the open domain is lacking to treat the restricted domains and all question types, because no restriction is imposed either on the user’s special vocabulary or on the question type, and it is very hard to construct a common knowledge (ontology) base for open domain. From the viewpoint of practicality, many researchers also begin to focus their attentions on the restricted domain QA and have built some advanced QA systems for restricted domain. On the other hand, since Chinese text retrieval has just been developed lately and there are many various specific characteristics in Chinese language, the research of Chinese QA using natural language was developed later than that in western countries and Japan. But, it has been started to pay attention to Chinese Question Answering in recent years. In this thesis, a fundamental start originates from the usability and limitation