Uma arquitetura para sistemas de recomendação de música baseado em contexto da interação e experiência do usuário

Abstract

This doctoral thesis addresses the research problem: "How can music recommendation systems be improved to more effectively integrate interaction context and user experience?". In response to this question, the research develops and evaluates UConteXt Arch, an architecture for music recommendation systems. This architecture is designed to enhance the user experience, explicitly considering the interaction context and musical preferences. The methodological approach adopted is Design Science Research (DSR), which enables iterative and cyclical analysis in the development of UConteXt Arch. The architecture is based on theories and research into recommendation systems, user experience, and contextual personalization. The development of UConteXt Arch involved a systematic literature review and exploratory studies of current practices on platforms such as Spotify and Deezer. The evaluation of UConteXt Arch included a variety of methods, such as the Communicability Assessment Method (MAC), the Technology Acceptance Model (TAM), UX Curves, and the Method of Intermediate Semiotic Inspection (MISI). These methods confirmed the effectiveness of the architecture in delivering accurate and adaptable music recommendations in line with changing user preferences and contexts. The architecture also addresses the ``cold-start'' challenge by implementing immediate feedback mechanisms and continuous learning to improve recommendations. The thesis concludes that UConteXt Arch represents a solution to the identified problem, demonstrating that it can significantly improve music recommendation systems by more effectively integrating interaction context and user experience. This study helps improve the quality of music recommendations and enriches the user experience, making a substantial contribution to the field of music recommendation systems.