IJMLC 2018 Vol.8(5): 466-471 ISSN: 2010-3700
DOI: 10.18178/ijmlc.2018.8.5.730

Conceptualization of Entity Relationship Based on Knowledge Graph

Yang Yu, Youlang Ji, Jun Zhu, Hongying Zhao, and Jingjing Gu

Abstract—As numbers of high-quality, large volume knowledge graphs appear, information extraction work has been enriched with more semantic knowledge. However, the entity relation extraction based on the knowledge graph is still at a very intuitive early stage, and the key issue it faces is the relation recognition and classification. In order to break the shackles of ontology and describe the relationship between the entities with fine-grained types, we propose a two-step bottom-up abstraction approach for relation conceptualization based on conceptual taxonomy that is automatically constructed. Given an entity relation, we figure out a group of Top-K concept pairs to abstract the relation, according to the typicality, diversity and coverage features. Our experimental evaluation shows that our method performing significantly high precision and quality for detecting fine-grained relationships.

Index Terms—Knowledge graph, entity relationship, conceptualization, clustering.

Yang Yu, Youlang Ji, Jun Zhu, Hongying Zhao are with the Jiangsu Electric Power Company, GaoYou County Electric Power Supply, Company Gaoyou 225600, China (e-mail: 642826764@qq.com).
Jingjing Gu is with the Jiangsu Electric Power Company, Jurong County Electric Power Supply Company, Jurong 212400, China.

[PDF]

Cite: Yang Yu, Youlang Ji, Jun Zhu, Hongying Zhao, and Jingjing Gu, "Conceptualization of Entity Relationship Based on Knowledge Graph," International Journal of Machine Learning and Computing vol. 8, no. 5, pp. 466-471, 2018.

General Information

  • ISSN: 2010-3700 (Online)
  • Abbreviated Title: Int. J. Mach. Learn. Comput.
  • Frequency: Bimonthly
  • DOI: 10.18178/IJMLC
  • Editor-in-Chief: Dr. Lin Huang
  • Executive Editor:  Ms. Cherry L. Chen
  • Abstracing/Indexing: Scopus (since 2017), EI (INSPEC, IET), Google Scholar, Crossref, ProQuest, Electronic Journals Library.
  • E-mail: ijmlc@ejournal.net