Designing an Architecture for Knowledge Graph Completion based on Discussion Agents using IBIS Structure

Abstract

Knowledge Graph Completion Challenge 2018 was held in order to approach to establishing explainable and interpretable artificial intelligence. We designed an architecture for knowledge graph completion based on discussion agents using the structure of IBIS (issue-based information system). The proposed architecture is an extension of our previous works, i.e., systems for supporting consensus building and agents for facilitating discussion on the Web. The IBIS structure is suitable for managing questions, hypotheses, and their evidences because these elements can be represented by nodes in an IBIS structure. Each hypothesis can be explained and evaluated on the basis of descendant nodes.