GA based CBR approach in Q&A system

Q&A (Question and Answer) is an important aiding tool for people to obtain information from the Internet. Since the existed question and answer engine maintains the feature weights gradually according to the users' response from the beginning, the prediction accuracy is too limited, and the maintaining process is boring and not efficient. In order to improve the prediction accuracy, in this paper, we introduce Genetic algorithm into traditional Q&A system and put forward a new architecture with Q&A engine which uses the conception of case based reasoning. The experimental results show that its prediction accuracy is greatly improved by our GA-based engine than other engines. (C) 2003 Elsevier Ltd. All rights reserved.