基于循环知识图谱的虚拟社区知识动态推荐方法

Dynamic recommendation method of virtual community knowledge based on circular knowledge map

针对数字化社区中资源数量庞大、用户难以第一时间方便且快速地寻找到所需数据的缺点,提出了一种基于循环知识图谱的虚拟社区知识动态推荐方法。将虚拟社区中知识点邻域实体当成上下文,获取知识表达学习实体,使循环知识图谱与待推荐数据相结合,计算用户历史点击信息,提取出实体特征向量;同时建立虚拟社区中用户模型,通过若干神经的协同过滤层、隐式交互用户以及知识点关系,实现多次非线性变换待推荐用户的隐式向量以及知识点隐式向量,完成动态推荐。实验证明:本文方法推荐满意度较高且推荐结果全面、不单一化。

For the disadvantage that the number of resources in digital communities is huge and it is difficult for users to find the required data easily and quickly at the first time,a dynamic recommendation method for virtual community knowledge was proposed based on cyclic knowledge graphs.The knowledge point neighborhood entities in the virtual community were treated as contexts to obtain knowledge expression learning entities.The cyclic knowledge graph was fused with the data to be recommended,the user′s historical click information was calculated,and the entity feature vector was extracted.Meanwhile,the user model in the virtual community with several neural co-filtering layers was established,implicitly interacting users and knowledge point relationships.The implicit vector of users to be recommended and the implicit vector of knowledge points were nonlinearly transformed several times to complete dynamic recommendation.The experiments prove that the recommendation satisfaction is high,and the recommendation results are comprehensive and not homogenized.