![]() ![]() ![]() There are several large song datasets that have been made available by researchers, most notably the Million Song Dataset. I chose to hone in on the concept of mood profiles for my project - could we use the predictions to create a mood profile for a potential streaming music service user and accurately represent their tastes?įeel free to check out my code on my github!ĭetermining how to compile a database of songs with mood labels presented a unique challenge. The music industry: Mood classification is increasingly becoming its bread and butter identification of moods can help with music recommendations, label/artist management, assembling music metadata and creating album, artist or playlist mood profiles. Mood classification can help to determine what kinds of songs evoke a brand’s image, and help create atmosphere for spaces and events. Some possible use cases of these predictions fall into two categories:Īssociation with music: For many consumer facing companies (e.g., retail, restaurants), music is part of a brand’s identity. There are many reasons why there would be a business case for this sort of problem - after all, moods and emotions are often what drive consumer purchasing decisions, which would be well within a record label, streaming provider or brand manager’s purview. With individual songs tied to mood labels, it seemed feasible to be able to use audio and track features to predict the mood of a song. I was interested in predicting how music evokes feelings, as the common denominator of shared experience for all music consumers. It enhances our activities, whether it be work, play or rest. Music accompanies us on every aspect of our lives. But mostly, it was about facing the music… My third project at Metis consisted of diving into the wonderful worlds of APIs and classification metrics.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |