When it comes to song recommendations, the basic algorithm is that it looks at the songs and genres and artists that you’ve been listening to the most, and comes up with recommendations for other songs/artists that are in a similar genre. However it isn’t always the most accurate because while an artist might be in one particular genre, your favorite song from that artist might be different from what the rest of the genre offers.

This is an area that Deezer thinks it can improve upon, and is turning to the use of AI (via Engadget) in which it can listen to a song and judge it by gauging its mood, emotions, and intensity, to help come up with smarter suggestions. It also uses both the music and the lyrics of the song itself to gauge what kind of music it is.

Deezer trained its AI to use raw audio signals and linguistic reconstruction models and a Million Song Dataset that aggregates Last.fm tags that describes tunes. Unfortunately it seems that this particular system isn’t quite ready for prime time just yet so don’t expect to see it deployed in Deezer’s services anytime soon, but it does hold a lot of potential and like we said, could eventually result in smarter playlists and song recommendations.

Filed in Audio >General. Read more about , , and .

Discover more from Ubergizmo

Subscribe now to keep reading and get access to the full archive.

Continue reading