Recommendation Algorithm Integrating Knowledge Graph and Sentiment Analysis

Against the backdrop of the explosive growth of digital reading resources, traditional recommendation algorithms are unable to effectively capture users' emotional preferences and are difficult to meet personalized demands. In the algorithm construction of this study, T-GCN was used to extract the features of KG to deeply explore the semantic associations and structural information among entities. The TF-IDF is utilized for feature extraction of sentiment analysis to accurately obtain the emotional tendencies of users. The features extracted from both are deeply integrated with the algorithm to achieve the dual drive of semantics and emotion. To verify the effectiveness of the algorithm, experiments were conducted using the Book-Crossing dataset and compared with the traditional models in the same environment. The results show that this algorithm is superior to the comparison algorithms, can effectively alleviate the problem of data sparsity, and improve the relevance and diversity of the recommendation results. This research provides new technical ideas for the recommendation field and has important theoretical and practical significance for promoting the development of the digital reading industry.