Automatic Question-answering System Based On Search Engines

Auto Question-Answering system gets more and more attention as a new way to obtain information, for it can accurately answer simple natural language questions. Conventional Auto Question-Answering system tends to use structured or semi-structured knowledge database system, thus there are disadvantages of limited areas and later updating. The advent of the Internet provides a new opportunity to re-built knowledge base for Auto Question-Answering system.Based on the Conventional search engine background, this paper puts out and implements a high-powered Chinese language search engine. It does a detail discussing about two key technologies in Auto Question-Answering system ---Question sentence type analysis and answer excavating algorithm by relying on the Chinese participle and famous entity etc. according to different question sentence types, characteristics and practical applications, it puts forward answers for each question sentence type and excavated algorithms separately. Meanwhile, this text realizes a common application system to support three kinds of questions, personage, time, and the place, which has actual practical value and expansibility. In this text, answer excavate algorithm has certain reference meanings to study Chinese information processing and common Auto Question-Answering system.The experimental result proves that Question sentence type analysis and answer excavating algorithm effectively improve MRR appraising value, with MRR 0.75 in the question of the personage type and average MRR 0.6 in three kinds of question sentence types. It can basically meet practical demand.Auto Question-Answering system is a very challenging subject. This text emphatically solves the problems of three kinds of basic question sentence types. It will make great efforts to study the settlement of the common type problem, repeated question sentences, relevance analysis, and further improvement of answer accuracy and the real-time performance of the system.