知识图谱的用户兴趣向量化方法及应用

User interest vector method based on knowledge graph and its application

为用户推荐其感兴趣的新闻内容,已成为了各大互联网新闻平台的首要技术目标。传统的新闻推荐方法主要是基于用户间的相似度或新闻内容间的相似度产生推荐列表。以上两种推荐方法虽然有效,但却忽略了新闻文本中存在的语义信息。知识图谱是一种描述实体以及实体之间链接关系的语义网络,基于知识图谱实现精准推荐是推荐系统目前的研究热点。本文基于知识图谱提出了一种用户兴趣向量的计算方法,在此基础上结合先进的卷积神经网络来构建推荐模型。所提出的基于知识图谱的新闻推荐方法,能借助知识图谱提取新闻文本中的部分语义信息,并将其应用于计算用户的兴趣向量,从而产生较好的符合用户语义的推荐结果。

Recommending news content of interest to users has become the primary technical goal of Internet news platforms.The traditional news recommendation method mainly generates a recommendation list based on the similarity between users or contents.Although the above two recommended methods are effective,they ignore the semantic information in the news text.Knowledge graph is a kind of semantic network describing entities and relationships between entities.It is a current research hot spot in recommendation systems to implement accurate recommendation based on knowledge graph.This paper proposes a calculation method of user interest vector based on knowledge graph,and then combines it with convolutional neural network to build a recommendation model.The proposed knowledge graph based news recommendation method can use knowledge graph to extract part of the semantic information in news text and apply it to calculate the user’s interest vector,so as to produce a recommendation result that better matches the user’s semantics.