The behavior of musicophiles has changed along with the evolution of the music industry in the past decades. Previously we conservatively bought music on a compact disc, but now music streaming services are more preferable; such as Amazon Music, Apple Music, Google Play Music, Pandora, Spotify, Youtube Music, to name a few. This is because of the convenience offered by these platforms so that users can search their favorite songs right away without having to bother going to the music store physically.
Users may not have enough time to scan through all available songs and manually create a playlist. Instead, a recommender system is constructed, which eases them to find relevant songs quickly. One example you might have seen before is the “Made For You” feature from Spotify. These personalized playlists are being recommended by grouping similar songs that go well together. How? In the real case, this process is done by combining several recommender algorithms based on users’ activities such as likes, playlist history, or even listening history.
This 3-day online workshop is beginner-friendly to demonstrate how to extract song embeddings using the neural network approach, specifically the word2vec model, to generate song recommendations. Throughout the online course, we will provide participants with hands-on examples and a rich interactive experience. One Instructor and two Teaching Assistants will help participants troubleshoot or help with any difficulties encountered.
NOTE: This workshop will be delivered in Bahasa Indonesia.