2002 Volume 14 Issue 5 Pages 461-471
This paper proposes an utterance generation model for systems that provide information via spoken dialogues. The information source is web sites on the Internet. The role of utterances in such systems is to provide information to users in an intuitively understandable form. Since the information on web sites is uncertain and inconsistent, the system has to acquire and integrate information from multiple sources in order to improve the reliability of information and generate utterances reflecting the reliability of the information. In our model, utterances are generated through two phases. The first phase integrates information acquired from multiple sources. The second phase generates linguistic expressions that reflect the reliability of integrated information. In this paper, we use the example of a weather forecast information system. This system gathers and integrates information from several weather forecast sites and then answers a user's question via dialogue.