基于概率推理的知识图谱中实体间关联度计算

Discovering entities correlation in knowledge graph based on probabilistic inferences

针对知识图谱中实体间的关联关系存在不确定性、实体间关联度计算复杂度高等问题,提出一种基于贝叶斯网的实体间关联度的计算方法。针对知识图谱做预处理,利用剪枝后获取的核心子图构建贝叶斯网,提出基于知识图谱的贝叶斯网构建方法;利用贝叶斯网作为知识图谱中实体之间关联关系的量化和推理框架,基于贝叶斯网的概率推理,提出知识图谱中实体间关联度的定量计算方法。建立在真实数据之上的实验结果验证了方法的有效性。

Aiming at the problems of uncertainty in the relationship between entities and high computational complexity of the correlation calculation between entities in knowledge graph(KG),a method of calculating the correlation between entities based on Bayesian network(BN)was proposed.The core subgraph obtained after pruning of the preprocessed KG,was used to construct BN based on KG(KBN).A method of KBN construction was proposed.BN was adopted as the quantization and infe-rences framework of the correlation between entities in KG,and based on the probabilistic inferences of KBN,a quantitative calculation method for the correlation between entities in the KG was proposed.Experimental results established on real data verify the effectiveness of the proposed method.