A Dialogue System for Collecting User’s Preference about Foods by focusing on Taste and Texture

Abstract

This paper proposed a dialogue system that collects user’s preference about foods by focusing on expressions of taste and texture. First, using a large recipe database, we constructed frame-based default knowledge base about dishes and their ingredients, and extracted co-occurring taste and texture expressions from Twitter. The dialogue system asks about user’s preference about foods based on the default food knowledge and dynamically determines the next question according to the user’s answer. We visualized the taste, texture, dishes and their ingredients as an example of utilizing acquired user’s preference.