Recommendation System Based on Interaction with Multiple Agents for Users with Vague Intention

摘要

We propose an agent-based recommendation system interface for users with vague intention based on interaction with multiple character agents, which are talking each other about their recommendations. This interface aims the user to make his/her intentions and/or potential opinions clear with hearing agents' conversation about recommendations. Whenever the user hits on any opinion, he/she can naturally join the conversation for getting more favorite recommendation. According to the result of experimental evaluation, the system with proposed interface can introduce more recommendations without any additional frustrations than the conventional recommendation systems with single agent.