Research Of Collaboratie Filtering Recommendation Algorithm Based On Domain Knowledge Map Entity Disambiguation

With the rapid development of the Internet today,people's lives are filled with more and more choices,so the recommended functions of major websites have emerged.When a friend recommends a book,visit the book website to find it,the target book appears in the return form,and the page will list some books you might like.This is the application of the recommendation system.For the traditional collaborative filtering recommendation algorithm,only the past opinions and behaviors of the original participants are used for prediction.This paper combines knowledge map technology to build a knowledge domain of book domain,disambiguation processing for entity semantics,semantic similarity calculation for entities obtained after entity disambiguation processing,and recommendation of semantic similar neighborhood and traditional collaborative filtering recommendation algorithm.The neighborhoods are fused according to different proportions,so that the semantic information about the Chinese entities is merged into the traditional collaborative filtering recommendation algorithm.The purpose is to improve the drawbacks of the traditional collaborative filtering algorithm without adding semantic information of the entity.The theory and experiment prove that the improved algorithm can effectively improve the effect of the collaborative filtering algorithm in the recommendation process,and improve the cold start problem in the recommendation process as much as possible.This paper proposes a collaborative filtering recommendation algorithm based on domain knowledge graph entity elimination,builds a domain knowledge base and constructs a knowledge map model through the book text information obtained by web crawler technology.The semantic similarity calculation is performed on the entities in the domain through the graph-based random walk algorithm,and the semantic similarity is established by establishing the model by using the text keywords of the unprocessed entities and the entity candidate keywords in the knowledge base.The cross-calculation selects the candidate entity with the largest calculated value as the selected target to realize the operation processing of the entity elimination.Finally,the semantic similarity matrix is constructed for the Chinese semantic data after the entity is processed and the semantic neighborhood is constructed.Finally,the semantic neighborhood and the neighborhood calculated by the recommendation algorithm are merged according to the corresponding experimental proportions,and the semantic neighborhood and traditional collaboration are adjusted.The filtering algorithm recommends the fusion ratio of the neighborhood and the domain knowledge map fusion dimension to train,and finally realizes the semantic level of the entity to be integrated into the recommendation field.This paper builds the knowledge base by using the entity information under the book category in the Wikipedia database,and uses the entity in the knowledge base and the book text information obtained through the web crawler technology.The test set is a good book list for crawling recommended reading on the book website,through Word2Vec The tool builds a knowledge map and a word vector model.The experimental evaluation results show the recall rate and accuracy rate and the F value as the reference value of the experimental results.The experiment proves that the algorithm can improve the effectiveness of the traditional collaborative filtering recommendation to a certain extent.