Improving Information Extraction with DPO for Dialogue-based Tourist-Spot Recommendation

Poster presentation, Dialogue System Symposium 2024.

We use Direct Preference Optimization to improve the information-extraction component of a CRS targeting tourist-spot recommendation, more reliably surfacing the user’s implicit preferences from natural dialogue.

J-GLOBAL entry

Recommended citation: Manato Tajiri, Michimasa Inaba. (2024). "Improving Information Extraction with DPO for Dialogue-based Tourist-Spot Recommendation." Dialogue System Symposium 2024 (poster).
Download Paper