Realization of the system that reconstructs information into a conversation

Abstract

Due to recent technological progress, various kinds of information, for example, objective information such as news and subjective information such as recommendations, are conveyed by various methods such as characters, photos, videos, robots, etc. Among them, it has become clear that conveying information using conversation makes it easier for the receiver to remain in memory than simply conveyed information. Therefore, by reconstructing information into conversation, we can convey information more efficiently. Two agents which has incomplete knowledge reconcile each other's knowledge and represent an attempt to understand each other's knowledge with conversation. Then we can reconstruct information into conversation. The two knowledge structures are created by dividing one complete knowledge structure into two incomplete ones. And the two agents aim to bring each knowledge structure closer to the complete one by reconciling each other's knowledge. The original complete knowledge structure is a structure with attributes, and form filling is performed using API. In addition, concrete conversations are created by filling incomplete sentences with words based on attributes. In this paper, we describe in detail the methods and techniques used for a system that realizes reconstructing information into conversation.