Emotion-Driven Deep Learning Recommendation Systems: Mining Preferences from User Reviews and Predicting Scores

Abstract

This paper presents a novel approach to recommendation systems by integrating emotion analysis from user reviews with deep learning techniques. We propose an Emotion-Driven Deep Learning Recommendation System (ED-DLRS) that mines user preferences and predicts scores by leveraging both the semantic content and emotional context of reviews. Our framework incorporates a dual-perspective emotion modeling strategy, considering both global emotion influence across the user base and localized emotional patterns of individual users. We introduce a deep neural network architecture that effectively fuses these emotion representations with latent user and item features. Extensive experiments on real-world datasets demonstrate that ED-DLRS significantly outperforms state-of-the-art recommendation methods, particularly in addressing the cold-start problem and data sparsity issues. Our results show an average improvement of 12% in prediction accuracy and a 15% increase in recommendation relevance compared to baseline models. Furthermore, we provide insights into the impact of different types of emotions on recommendation quality and user satisfaction. This work opens new avenues for emotion-aware, personalized recommendation systems that can enhance user experience in e-commerce and content delivery platforms