Ji-Hun Hong  Hyun-Young Choi  Wan-Gon Lee  Young-Tack Park

10.5626/JOK.2019.46.2.131

Knowledge Base knowledge completion Neural Network Embedding

Recently, with the increase of the amount of information due to the development of the Internet, there has been an increased interest in research using a large-capacity knowledge base. Additionally, studies are being conducted to complete the knowledge base as it uses become widely used in various studies. However, there has been lack of research to detect error triples in the knowledge base. This paper, we proposes the embedding of an algorithm to detect the error triple in the knowledge base, the utilization of the clustered embedding model and the four relational models, which are typical algorithms of triple classification. Additionally, a relation ensemble model was generated using the results of the single embedding models and the embedding ensemble model similarly generated using the results of the single embedding models. The error triple detection results were then compared and measured through the model verification indexes.