A proposal of topic-map-based QA system for natural language input

Abstract

In this paper, we propose a QA system that returns a suitable answer to a question submitted by a user. It divides a question sentence into units, each of which is a tuple of a noun, a particle, and a predicate. The QA system returns the answer whose assigned units match the units of the question. In order to absorb the variant of particle usage, we applied cluster analysis to questions in Yahoo! Answers to group particles. In order to maintain the relationships between units and answers, we utilized topic maps.