2 ms·
I've never found a perfect recommendation for me simply because data-driven approaches are 1) backwards looking and 2) assume I have a stable set of preferences
by mushufasa 7y ago
I've never found a perfect recommendation for me simply because data-driven approaches are 1) backwards looking and 2) assume I have a stable set of preferences. My preferences change every month, are tied to certain activities (e.g. workout), and based on mood.
So now I mostly listen to NTS. If I don't like what's on the air I search for a recent episode that looks good. I highly recommend it. https://nts.live https://nts.live.
- jacobobryant 7y agoI'll check nts out. Although it's not super developed yet, I do intend for lagukan to handle unstable preferences. It already kind of handles different sets of songs--i.e. it doesn't think "You like song A," it thinks "You just listened to song A, and previously when you listened to A you also listened to B". So it can adapt based on the songs in your current listening session, although it takes a while for it to learn the different sets. I'm planning later to make it adapt to preferences changing over time and also contextual factors like working out. There's a lot of good research on context-based music recommendation.
- xwowsersx 7y agoLoving NTS, thanks. How do I get the artist/title of the track I'm listening to? Really want to save some of this.
- mushufasa 7y agoMost shows display their tracklist e.g. https://www.nts.live/shows/ruf-dug/episodes/ruf-dug-28th-june-2019 https://www.nts.live/shows/ruf-dug/episodes/ruf-dug-28th-jun... scroll down a bit. Often the hosts talk about the tracks.