ISCA Archive Interspeech 2025
ISCA Archive Interspeech 2025

Improving User Impression of Spoken Dialogue Systems by Controlling Para-linguistic Expression Based on Intimacy

Shoki Kawanishi, Akinori Ito, Yuya Chiba, Takashi Nose

The advent of large language models (LLMs) has improved the naturalness of responses in dialogue systems; however, conversations with these systems still differ significantly from those between humans. For instance, humans adjust their manner of speaking based on their relationship with the conversation partner, whereas dialogue systems respond uniformly, even after repeated interactions. By producing responses that consider the progression of relationships, it may be possible to develop dialogue systems that are more appealing to users. Previous studies attempting to achieve such systems have primarily focused on the linguistic expressions of responses,while paralinguistic expressions have been largely overlooked. In this paper, we propose a dialogue system that adapts both linguistic and para-linguistic expressions as the number of interactions increases. We also evaluate its effectiveness through dialogue experiments.