A Heterogeneous Learning Resources Recommendation Algorithm Combination of Content-Based and Keyword Vectors Weight

Abstract

At present, there are a variety of learning resources which can be applied to personalized recommendations in the field of online education. While being stored, these learning resources should be described in text. Therefore, the use of content-based recommendation algorithms in the resource recommendation field can achieve good recommendation results and can solve the cold start problem of collaborative filtering. However, because traditional content-based recommendation algorithms only consider the statistical characteristics of text when modeling items, it has certain semantic analysis defects and cannot handle unstructured data such as audio and video. In this paper, first, aiming at the problem that there are many types of learning resources and the structure are quite different, the resource standardization metadata structure and heterogeneous resource mapping rules are defined, then build a standardized modeling framework for heterogeneous resources and realize the standardization of heterogeneous resources. Subsequently, the problem of semantic analysis flaws in traditional content-based recommendation algorithms can be solved by combining the traditional content-based recommendation algorithm and the Word2Vec word vector model with semantic analysis advantages, the statistical features and semantic features of the resource text are comprehensively extracted to perform resource feature modeling to solve the problem. The traditional semantic analysis defect problem based on content recommendation algorithm is solved, and the similarity calculation method of traditional algorithm is improved to further improve the recommendation effect of recommendation algorithm.