I have not started my research yet, so the following is a plan of my study.
The goal of this research is to develop a question answering (QA) system which answers user’s question by searching on the Internet. In order to accomplish this goal, the behaviour of Community Question Answering (CQA) websites (such as Yahoo! Answers) will be analysed.
There are plenty of sample pairs of questions and answers on CQA websites. There, we can find that, even if two questions are almost the same, their answers may be different depending on the information need of the person who asks the question. For example, answers for the question “how can I reply to the email?” vary with who sends it and what it says. Traditional information retrieval systems often fail to find the information which meets user’s needs. In this research, we will assume the answers of a question and use this assumption to retrieve information on the Web. Although it may sound circuitous, we believe it is a useful way because the amount of information on the Web is far larger than that on CQA sites.
This QA system will make a database by tagging instances of questions in CQA sites, those of answers, and webpages on the Intenet in advance independently. Then, it will also tag user's query and match the tags to the tags of the database. Main task of my research will be to find the method to tag them effectively and efficiently.