摘要： 在马尔科夫逻辑网(MLN)的实体解析算法中任意闭原子采用硬约束,导致推理及权重学习过程较难收敛到最优解,降低解析精度及执行效率。为此,提出一种将概率软逻辑(PSL)模型应用到实体解析中的方法,该模型中闭原子采用软约束,易于进行知识推理与权重学习。阐述PSL模型基本理论,通过实体关系、实体属性、本体约束构造PSL模型的逻辑规则,描述实体解析的匹配过程,根据PSL模型的推理机制实现实体解析的决策过程。实验结果表明,与基于MLN的实体解析算法相比,该方法可大幅提高实体解析的准确率、F1值及执行效率。

关键词: 实体解析, 概率软逻辑, 马尔科夫逻辑网, 实体关系, 实体属性, 本体约束

Abstract: As any closed atom adopts hard constraints in Entity Resolution(ER) algorithm based on Markov Logic Network(MLN),the reasoning and weight learning processes of the algorithm hardly converge to the optimal solution,which decreases the efficiency and accuracy.This paper proposes a method to apply Probabilistic Soft Logic(PSL) model to entity resolution,the closed atom in the model adopts soft constraints,making it easy to carry out millions of knowledge reasoning and weight learning.The paper explains on the basic theory of PSL model,and constructs a first order logic rule to describe the matching process of entity parsing through three aspect of the entity relationship,entity attribute,and ontology constraints.The reasoning mechanism is used to calculate the entity matching result accurately and efficiently.Experimental results show that,compared with the physical analytic method based on MLN model,this method can improve the accuracy and efficiency of F1,and significantly improve the execution efficiency.

Key words: Entity Resolution(ER), Probabilistic Soft Logic(PSL), Markov Logic Network(MLN), entity relationship, entity attribute, ontology constraints

中图分类号: 

• TP391