QUESTION ANSWERING FROM FREQUENTLY ASKED QUESTION FILES - EXPERIENCESWITH THE FAQ FINDER SYSTEM

This article describes FAQ FINDER, a natural language question-answeri ng system that uses files of frequently asked questions as its knowled ge base. Unlike Al question-answering systems that focus on the genera tion of new answers, FAQ FINDER retrieves existing ones found in frequ ently asked question files. Unlike information-retrieval approaches th at rely on a purely lexical metric of similarity between query and doc ument, FAQ FINDER uses a semantic knowledge base (WORDNET) to improve its ability to match question and answer. We include results from an e valuation of the system's performance and show that a combination of s emantic and statistical techniques works better than any single approa ch.