Int J Performability Eng ›› 2017, Vol. 13 ›› Issue (8): 1293-1303.doi: 10.23940/ijpe.17.08.p11.12931303

• Original articles • Previous Articles     Next Articles

Entity Disambiguation with Markov Logic Network Knowledge Graphs

Jiangtao Ma, Tao Wei, Yaqiong Qiao, Yongzhong Huang, Weibo Xie, Chaoqin Zhang, Yanjun Wang, and Rui Zhang   

  1. aState Key Laboratory of Mathematical Engineering and Advanced Computing, Zhengzhou 450002, China
    bZhengzhou University of Light Industry, Zhengzhou 450002, China
    cNational Digital Switching System Engineering & Technological R&D Center, Zhengzhou 451000, China
    dHenan Institute of Engineering, Computer College, Zhengzhou 451000, China
    eNorth China University of Water Resources and Electric Power, Zhengzhou 450002, China

Abstract: Disambiguating named entities is an important problem in natural language processing, knowledge base, question answering systems. In the paper, we propose a Markov logic network knowledge graph solution for solving entity resolution problem. First, we employ knowledge graph to represent the entity relationship between linked entities in the knowledge base. Then, we utilize MLN to inference the inconsistent relationship in the knowledge graph, and disambiguate the entities in the process of entity disambiguation. As far as we know, inferencing with MLN is a first attempt for entity disambiguation in the knowledge graph. We evaluate the proposed solution with three real world knowledge bases and compare it with four baseline solutions. The experimental results demonstrate that our solution is 7% higher than other baseline methods with F1 measure. We also test our scheme and compare entity resolution systems on four datasets with three knowledge base corpora. Extensive experiments show that our solution achieves higher precision and recall than baseline solutions.


Submitted on July 25, 2017; Revised on August 30, 2017; Accepted on September 15, 2017(This paper was presented at the Third International Symposium on System and Software Reliability.
References: 30