Automatic question answering based on large text knowledge base

摘要

The best way to find out something is to ask an expert, because the expert can clarify what the questioner really wants to know through the conversation. This paper discusses how to construct such an expert-like question answering system, more specifically, how to ask back for an unclear question. The basic idea is to detect the difference between the retrieved texts to generate asking back expressions. We introduce two dialogue systems: a dialogue helpsystem for Kyoto University Computer Center, and a dialogue helpsystem for Microsoft products.