基于改进贝叶斯的知识图谱关联查询算法仿真  

Simulation of Knowledge Graph Correlation Query Algorithm Based on Improved Bayes

作　　者：陈凌 火明刚 陶雪娇 朱长娥 CHEN Ling;HUO Ming-gang;TAO Xue-jiao;ZHU Chang-e(School of Software and Artificial Intelligence,Chongqing Institute of Engineering,Chongqing 400056,China;Lanzhou Jiaotong University School of Mathematics and Physics,Lanzhou Gansu 730000,China)

机构地区：[1]重庆工程学院软件与人工智能学院,重庆400056 [2]兰州交通大学数理学院,甘肃兰州730000

出　　处：《计算机仿真》2023年第7期453-456,485,共5页Computer Simulation

基　　金：重庆市教委科学技术研究计划项目(KJQN202101903)。

摘　　要：知识图谱的不确定性导致节点之间的语义关联性不高,抽取出的文本存在较多冗余和噪声,无法准确捕捉用户的查询意图。为解决上述问题,提出改进贝叶斯框架的知识图谱关联查询算法。通过将知识图谱中的文本输入深度卷积神经网络,编码处理获得文本表示,并在非对称映射的基础上获得结构表示,以此获得知识图谱关系与实体之间的关系。采用低秩矩阵映射补全知识图谱中的尾实体和头实体。优化贝叶斯推理框架对实体量化处理,挖掘知识图谱中存在的数据集,并计算关系关联度和属性关联度,根据得分排序查询结果,完成知识图谱的关联查询。实验结果表明,所提算法的F1值在0.8以上,且准确率和召回率高,说明具有良好的查询性能。The uncertainty of the knowledge graph may lead to low semantic relevance between nodes,and the extracted text contains many redundant information and noise points.In this paper,a knowledge graph association query algorithm based on improved Bayesian framework was proposed.Firstly,the text in the knowledge graph was input into the depth convolution neural network,and then the text representation was obtained by encoding.Meanwhile,the structure representation was obtained based on asymmetric mapping.On this basis,the relationship between knowledge graphs and entities.Low-rank matrix mapping was used to complement the tail entity and head entity in the knowledge map.Moreover,the Bayesian reasoning framework was optimized to quantify the entity and mine the data set from the knowledge graph.Finally,the relationship correlation degree and the attribute correlation degree were calculated.According to the score,query results were sorted,thus completing the association query of the knowledge graph.Experimental results show that the F1 value of the proposed algorithm is more than 0.8,and the accuracy and recall rate are high as well,indicating that the algorithm has good query performance.

关 键 词：改进贝叶斯框架 深度卷积神经网络 知识图谱补全 知识图谱 关联查询算法 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]