Research Of Recommendation Algorithm Based On LDA And Word2Vec

With the rapid development of virtual equipment maintenance technology, research and development of interactive training aid system receives more and more attention. However, a large number of domestic and international virtual maintenance training systems require trainees to understand and master the system deeply, and they are not able to provide timely answers and help. Therefore, it is a serious problem of how to make trainees learning training model more efficient and convenient and have a better interactive experience. Aiming at the above problems, the paper researched the personalized questions and recommendations for users and completed the related system. The main achievements of this paper can be summarized as follows.(1) Completed analysis and research of Q&A system and personalized recommendation technology and its current situation, then further completed the study and improvement of related models.(2) Completed in-depth study on key technologies of the system and made experimental verification. First LDA topic model and its principles have been studied. The model can tap the potential topic distribution information of the text and we can get the knowledge document set and distribution of user documentation in different themes through training of LDA model. Then, we design experiments to compare LDA model with the traditional vector space model, and the results demonstrated the advantages of LDA model on the correlation of matching documents. Secondly, the word2vec and the word vector techniques were studied, and this paper vertify its good performance on text similarity calculation through experiment. (3) Based on the research and related experiments on LDA and word2vec, an improved similarity calculation algorithm was proposed, which combines the themes and deep semantic information and can in-depth tap the potential semantic information of text and make the similarity calculation more accurate. The algorithm has achieved good results through our actual experiments. Then, this paper aslo studied the technologies of personalized document recommendation and question recommendation for users and provided more accurate document and question recommendation for different users to help users achieve higher efficiency of getting information and better access to personalized interactive experience.(4) Based on the foregoing research and related work, we completed the overall architectural design, database design, process flow analysis, functional module design of the system. Then we achievied the development of the system with using the relevant technologies based on B/S structure.