Do you ever feel like Spotify knows your music preferences and moods better than you do?
Has Spotify introduced you to your new favorite artist? Has it customized your go-to playlist that you listen to on every vacation you take? Or did you find that new podcast you can’t stop recommending to your friends on Spotify? This isn’t by coincidence.
Our platform uses an artificial intelligence system called Bandits for Recommendations as Treatments (BaRT) that’s main purpose is to keep our listeners engaged. BaRT helps with this by suggesting and playing music that it knows will be familiar to a listener, while also sprinkling in new songs they’ve never heard before but might like. This AI software is responsible for your home screen, where you can see artist spotlights and custom-generated playlists, in addition to helping us generate the Release Radar, Daily Mix, and Discover Weekly features for every one of our users.
There are three main functions that BaRT uses to suggest songs to our listeners:
- Natural Language Processing: BaRT analyzes the language, lyrics, and contents of a song to see if it would be a good match.
- Raw Audio Analyzation: BaRT can detect the so-called ‘vibe’ or ‘mood’ of a song and then place it into a category (upbeat, chill, instrumental, heavy, etc.) before uploading it to a personalized playlist for someone.
- Collaborative Filtering: BaRT can compare new songs to a listener’s existing music preferences to decide what artists and songs will best compliment their tastes.
In addition to collecting data from all of the songs on our platform, we can also help tailor our algorithm to our listener’s tastes by collecting data from them. Attributes like gender, age, and location can help us give better suggestions to all of our Spotify listeners. We know data collection can often make certain users a little weary, but it really does contribute to our ability to give our listeners more accurate recommendations that will keep them happy.
Our algorithm can even look for patterns in how someone listens to music. How long do you play a certain song or artist? What songs do you typically skip? What kinds of playlists have you put together yourself? This information is more insightful than you might think, as it helps determine the suggestions that are given to you and you alone. Additionally, the algorithm also studies other users around the globe; if Spotify finds a user who is similar to you, that information can be used when putting together your custom-generated playlists.
Spotify is renowned for its ability to make spot-on recommendations. If you’ve ever wondered how that’s possible, I hope this blog has given you a little glimpse into how BaRT has made a difference in the way you listen to and enjoy your music. The impact that AI has on personalization is extraordinary, and we’re glad we can use it for the benefit of our listeners.