KGIPSL: A Knowledge Graph Inference Method based on Probabilistic Soft Logic

摘要

Knowledge graph inference has a wide range of applications in semantic search, question answering systems, entity disambiguation, link prediction, and recommendation systems. However, the accuracy and operational efficiency of existing methods do not meet the needs of large-scale knowledge graphs. Aiming at the problem of large-scale knowledge graph inference, this paper proposes a knowledge graph inference method based on probabilistic soft logic (KGIPSL). Firstly, KGIPSL uses the Markov logic network to construct the relationship between entities. Secondly, KGIPSL employs probabilistic soft logic to represent non-deterministic knowledge and infers the relationship between entities in the knowledge graph. Thirdly, KGIPSL conducts accurate knowledge inference. Experiments on real knowledge graph datasets show that the KGIPSL method is superior to the existing baseline method in accuracy, recall, and efficiency. Among them, the average accuracy of KGIPSL on the YAGO dataset is 14.9% higher than that of the baseline method.