MultiMoodIfy a Lego-Like Emotion-Aware Music Recommender for Spotify

Artificial Intelligence (AI) is gaining increasing space in daily life. Modern intelligent interfaces are fed by smart multimodal engines, which enable users to interact through multiple preferred or assistive channels to express their needs in the most intuitive, natural, and comfortable way. The new frontier of these agents includes, among others, the use of sophisticated natural language processing algorithms, visual question-answering abilities, and the possibility to capture users’ intentions and anticipate their needs even when they are not explicitly expressed. The latter entails using advanced techniques such as emotion recognition by expression, voice, or physiological signals. This work shows how existing architectures can serve as blocks to build a multimodal recommender system for Spotify, which captures the user’s mood over different channels.