Research On Learning Resource Recommendation Based On Knowledge Graph And Collaborative Filtering

The accuracy of learning resource recommendation is the key to realize precision teaching and personalized learning.However,due to the massive data constantly generated in the Internet,it will lead to the emergence of “information overload”、“knowledge astray” and other problems,so it is difficult for learners to accurately locate the learning resources they need on the massive network learning platform,even more difficult to realize personalized learning.Collaborative filtering algorithm is a classic and commonly used recommendation algorithm,but it has a cold start issue.However,the knowledge graph contains rich semantic content,it can combine various types of information from different sources to form a complex network of relationships,allowing learners to analyze problems from the perspective of “relationships”.By using the knowledge graph,more reference information can be provided to the recommendation system to help it better meet the needs of users.In addition,the knowledge graph can also solve the cold start and data sparse problems of traditional algorithms.Therefore,this thesis proposes to combine knowledge graph and collaborative filtering algorithm to recommend learning resources,so as to improve the accuracy of learning resource recommendation.The research work of the thesis is as follows:1.The ALBERT-Bi LSTM-CRF named entity recognition model was built.Considering the ALBERT model occupies less memory and can identify entity names more accurately,the thesis uses ALBER model instead of BERT model.The ALBERT model can extract semantic information from text to enhance the semantic expression of text.In the thesis,the ALBERT model is used to obtain the dynamic word vector,and Bi LSTM network is used to realize depth feature learning.At the CRF layer,by analyzing the correlation between different categories of labels,the accuracy of recognition can be improved,thereby achieving entity recognition of text information.Finally,the effectiveness of the proposed model is verified by several groups of comparative experiments.2.Constructing the knowledge graph of learning resources fields.Firstly,the relevant data is obtained by Scrapy framework in Python,and the data is preprocessed by knowledge extraction and knowledge graph to obtain information in various fields and convert it into triplet form.Then,this information is stored in the Neo4 j graph database to build a learning resource knowledge graph that clearly shows each entity and its relationships in a graphical form.3.A learning resource recommendation model combining knowledge graph and collaborative filtering is proposed.This model not only uses the relevance within the course,but also uses the user behavior and other information to integrate the knowledge graph into the collaborative filtering recommendation algorithm,so as to improve the recommendation effect.Finally,the validity of the model is analyzed by comparing with other models.Experiments show that compared with the traditional collaborative filtering algorithm,the model proposed in the thesis improves the recall rate,accuracy and F value.