An Effective Similarity Measurement for FAQ Question Answering System

FAQ system is a QA retrieval system which finds the question sentence from question-answer collection and then returns its corresponding answer to user. The task of matching questions to corresponding question-answer pairs has become a major challenge in FAQ system. In this context, we address the problem by using a semantic knowledge base, namely, word co-occurrence corpus, to improve its ability to match question and answer. We propose a method for similarity measurement between questions according to the number of the relative terms as well as the length of question sentence. Experiments on real datasets give evidence that a combination of semantic and statistical techniques works better than any single approach.