Text mining deep hybrid recommendation model based on attention mechanism
ZHANG Jing1， CHEN Zengzhao2， DUAN Chao3， WANG Hu2
(1.School of Education， Jianghan University， Wuhan 430065，China；2.Faculty of Artificial Intelligence in Education， Central China Normal University， Wuhan 430079，China；3.Key Laboratory of Intelligent Education Technology and Application of Zhejiang Province， Zhejiang Normal University， Jinhua 321004， Zhejiang，China)
Abstract：At present， it is becoming more and more difficult to obtain video resources that meet the needs of users from massive internet information， and users are facing serious information anxiety and information overload problems. However， there is a large amount of information related to users' preferences and item characteristics， which has not been exploited in the classical recommendation systems. In view of the outstanding performance of deep learning in feature extraction and attention mechanism in feature selection， making full and effective use of various auxiliary information to alleviate contradictions is a hot and difficult issue in current research. In response to the above problems， a novel deep hybrid recommendation method using textual context information is proposed in this paper. This method combines the video title and video introduction， and converts pre-trained word embedding model Glove into word vectors. The convolution neural network combined with multi heads' attention mechanism is used to extract the hidden factors of items， and the probability matrix decomposition is used to predict the users' rating of video resources. The experimental results on four public datasets， ML-100k， ML-1m， ML-10m， and Amazon， show that the proposed method outperforms the baseline models such as PMF， CDL， and ConvMF.
2023, Vol. 57 
