JSAI2019

Presentation information

Interactive Session

[3Rin2] Interactive Session 1

Thu. Jun 6, 2019 10:30 AM - 12:10 PM Room R (Center area of 1F Exhibition hall)

10:30 AM - 12:10 PM

[3Rin2-28] Improvement of Knowledge Graph Completion Using Label Characters for Questions to Acquire Knowledge in Dialog Systems

〇Yuma Fujioka1, Katsuhiko Hayashi1, Mikio Nakano2, Kazunori Komatani1 (1. The Institute of Scientific and Industrial Research, Osaka University, 2. Honda Research Institute Japan Co.,Ltd.)

Keywords:Dialog System, Knowledge Graph Completion, Knowledge Extraction

Dialogue systems cannot respond about information that is not explicitly described in their knowledge bases. Constructing a perfect knowledge base is practically impossible; that is, filling all the values in databases is quite labor-intensive.
We are trying to construct a system that can acquire information that is not explicitly described in the knowledge base by inferring latent information from knowledge graphs.
In particular, we complement the links in a knowledge graph by using an embedding into latent space.
We use partial character sequences of labels (i.e. entity names) to improve of knowledge graph completion.
We also show examples of queries generated using the latent information in our target knowledge graph.