Abstract:
"In this research, the challenge of enhancing book recommendations is tackled by considering a user’s current emotional state and contextual information, such as age and location, which are often overlooked in traditional systems. Previous approaches primarily relied on collaborative and content-based filtering methods, which, while useful, did not account for the user's current mood or context, leading to a lack of personalized recommendations.
The methodology adopted in this research differs from conventional book recommendation
systems by integrating facial expression analysis with sentiment analysis of book abstracts,
introducing a unique emotional context to the recommendations. Utilizing the FER library, the system captures real-time emotional states of users, while a sentiment analysis model trained on BERT and enhanced with additional layer evaluates the emotional tone of book abstracts. Subsequently, content-based, and context-based filtering mechanisms are applied, utilizing the detected emotions along with contextual data such as age and location. This innovative approach is designed to synchronize book suggestions with the user’s current emotional state and personal circumstances, providing a deeply personalized and contextually aware reading experience.
Testing of the sentiment analysis component shows promising results, with high precision and recall for several emotions, and an overall accuracy rate of 93.75%, confirming the potential of this research's approach to personalized book recommendations. By integrating emotional intelligence with contextual information, the system offers recommendations that resonate with the user's current mood, age, and location, paving the way for a more intuitive and empathetic user experience."
