A Proposed Automatic Answering System for Natural Language Questions

Abstract

Though computers are considered as machines, they are expected to give answers to questions in normal English language, just the way human beings can do. To train a computer to answer English language questions is an interesting and challenging problem. The automatic answering systems are classified under two categories: open domain systems and close domain systems. In this paper an answering scheme is proposed that combines close domain and open domain system. The question answering system for closed domain deals with questions under specific domain. Answers to questions from close domain cannot be searched using a search engine. During a close domain, answers to questions are not alive in the public area and so they cannot be search using a search engine. Hence, the answers to such questions are maintained in a database by the domain expert. During retrieval of answers, the best matched answer searched from database is returned to the user. To perform this matching a template matching technique is used. Open domain question answering system deals with all questions covering all domains. These systems can rely search engines to find the required answer. In this paper, we have combined both the approaches and proposed an answering scheme which accepts input in General English language. Thus the main challenge of the proposed scheme is to understand the English language question.