Emotion-aware hybrid explainable movie recommender system (EHEMRS)

Abstract

Movie recommendation systems play a vital role in digital entertainment, helping users navigate vast content libraries to find movies that match their preferences. These systems enhance user engagement through tailored suggestions, addressing the challenge of overwhelming movie choices. Traditional approaches, often rely on historical ratings or simple content attributes, overlooking the emotional context shaping viewer preferences. This can lead to recommendations misaligned with a user’s current mood or psychological state. The Emotion-aware Hybrid Explainable Movie Recommender System (EHEMRS) tackles this issue by integrating emotional analysis with clear, user-focused explanations for highly personalized and reliable recommendations. EHEMRS combines collaborative filtering, content-based filtering, VADER sentiment analysis, NRCLex, and Plutchik’s wheel to capture not only the basic eight emotions from NRCLex but also complex emotional states, enabling deeper alignment with users’ emotional experiences during movie selection. LIME and SHAP methods further clarify the system’s decision-making process, enhancing users trust. Evaluated on IMDb and MovieLens datasets, EHEMRS achieved precision scores of 0.9354 and 0.9720, respectively, outperforming standard models like SVD and other emotion-aware systems such as Chat-Rec. Plutchik’s wheel helps the system capture a wider range of emotions, allowing the system to detect more subtle emotions like love or optimism, which are equally important to users’ watching experience. This method makes recommendations connect better with users’ feelings creating a more enjoyable watching experience. Future improvements should focus on improving computational efficiency to shorten training times and incorporate multilingual sources to minimize biases known in English-based tools like NRCLex.