This paper proposes a method with which an interview dialogue system can learn user-friendly dialogue strategies. Conventional interview dialogue systems mainly focus on collecting the user’s information and simply repeat questions. We have previously proposed a method for improving user impressions by engaging in small talk during interviews that performs frame-based dialogue management and generates small-talk utterances after the user answers the system’s questions. However, the utterance selection strategy in the method was fixed, making it difficult to give users a good impression of the system. This paper proposes a method for learning strategies for selecting system utterances based on a corpus of dialogues between human users and a text-based interview dialogue system in which each system utterance was evaluated by human annotators. This paper also reports the results of a user study that compared the proposed method with fixed utterance selection strategies.

Takahiro Kobori

Papers

Learning Dialogue Strategies for Interview Dialogue Systems that Can Engage in Small Talk.