Research And Application Of Recommendation Technology Based On Knowledge Graph

With the popularization of the Internet and the development of information technology,the network information data has shown an explosive growth,and the problem of information overload has been highlighted.In order to help users find information that they are interested in from a large amount of information,and to help information producers to make the information they produce come to the attention of a large number of users,the recommendation system came into being.However,the sparsity problem and the cold start problem faced by the traditional recommendation system have limited the effectiveness of the recommendation system to some extent.This paper constructs a domain knowledge graph,extracts the semantic representation of the items,and combines the collaborative filtering algorithm to improve the performance of the recommendation system.This paper takes the research and application of recommendation technology based on knowledge graph as the research topic,and focuses on the construction of the domin knowledge graph,the embedding method of entities and relationships in the knowledge graph,and the recommendation algorithm based on knowledge.The main research content of this article is as follows.Firstly,this paper studies the construction method of domain knowledge graph,designs and implements a method of constructing knowledge graph of the movie domain.This paper introduces the concept of knowledge graph into the field of movie,researches and analyzes the characteristics of the knowledge about movie,realizes the division of entities and relations in the field,and completes the construction of the movie ontology library.Then we extract the entities and relationships in the movie domain and complete the storage of knowledge maps based on a relational database.Finally,we analyze the advantages and disadvantages of the relational database for knowledge storage,and use Neo4 j to improve the storage of knowledge.Secondly,this paper presents and implements a vectorization method for entities and relationships in the domain knowledge graph.We present an embedding model of entities and relationships in the knowledge graph.The knowledge representation technology is used to vectorize the entities and relationships on the basis of preserving the semantic information.In addition,this paper also improves the original negative sampling algorithm.The experiment proves that the vectorization method used in this paper is effective.Finally,this paper proposes a personalized recommendation algorithm based on knowledge map.Combining the vectorized representation of the items in the knowledge map and the collaborative filtering algorithm,using the article semantic information extracted from the knowledge graph,it can make up for the defects that the item-based collaborative filtering algorithm does not consider of the content information of the item itself.Compared with the item-based collaborative filtering algorithm,the recommendation algorithm based on knowledge map presented in this paper has a certain improvement in precision,recall,and coverage.