Research And Implementation Of Course Recommendation Algorithm Fused With Semantic Information And Hierarchical Attention Network

With the continuous development of Internet technology and the rise of online education,the research on course recommendation algorithms has become more and more important.Online education platforms provide learners with a convenient learning path by providing a large number of learning resources.However,massive course resources also bring about the problem of information overload,and learners need to spend a lot of time and energy to select courses that meet their needs.In this case,personalized course recommendation algorithm becomes an effective way to solve this problem.Course recommendation algorithms can provide learners with more accurate and useful learning resource recommendations based on their individual needs and interests.Although the research work on recommendation algorithms has achieved good results in other fields,their applications in online course recommendation systems are relatively few.First of all,when using traditional course recommendation algorithms,there are problems such as data sparseness and cold start,and the semantic information in texts such as course introductions is not fully utilized;secondly,users’ interest preferences change over time,ignoring this problem leads to The recommendation results still need to be further improved.In this paper,we propose a course recommendation model that fuses semantic information and hierarchical attention networks to solve the above problems.In addition,this paper also designs and develops a course recommendation system that integrates semantic information and hierarchical attention networks to verify the applicability of the model.The specific research contents are as follows:(1)Aiming at the sparsity of online course data in the previous course recommendation,attribute feature information such as course introduction was ignored.This paper proposes a course recommendation model based on semantic information extraction.The model uses Text CNN to extract the semantic information in the course introduction,and obtains the course Embbeding with richer attribute characteristics;at the same time,according to the user’s course interaction session records,the attention network is used to mine the user’s interest weight in different courses,and the interaction between the user and the interactive course is obtained.Deep Feature Representation for Improving Recommendation Accuracy.(2)Aiming at the problem that the user’s dynamic interest preference changes over time,this paper uses a hierarchical attention network to simulate the user’s dynamic interest preference change,and makes full use of the semantic information of courses to make recommendations,and then proposes a fusion of semantic information and hierarchical attention.A course recommendation model for the web.The model uses Text CNN to extract semantic information to generate course embedding vectors;in order to capture the user’s dynamic interest preference,first use an attention network to capture the user’s long-term interest preference from the user’s learning history,and then use another attention network to The long-term interest preference is combined with the courses with different weights in the short-term session records to obtain the user’s final interest preference,and finally the two are fused and recommended according to their scores on courses.Experimental results show that the model can generate more effective recommendation results.(3)On the basis of the improvement of the above algorithms,a course recommendation system integrating semantic information and hierarchical attention network is designed and developed.The system can generate personalized recommendation results based on the user’s learning behavior records to improve the user’s learning efficiency;at the same time,the user behavior data set collected by the system also provides data support for the next research work.