Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Spotify’s Discover Weekly: How machine learning finds new music

hackernoon.com

Post reply on HN
Print view View original

31–40 of 275 posts

Re: Spotify’s Discover Weekly: How machine learning finds new music

#31

Post by Jdam »

chewz wrote: post #24

With all that AI excitement Spotify, Google Play Music and Apple Music misserably fail at generating good musical recomendations. At least for me. Spotify Discover Weekly was no different for me when I still used Spotify (six months ago). Not only it is not personalized at all it is also quite lame. I am doing enough to seed - following enough artists, genres, liking album, subscribing to playlists but the quality of…

My experience was that Spotify radio is utterly bad, Discover Weekly quite okay, but the weekly Apple Music playlist really nails it. I found a lot so far that I didn't know before and added to my library.

Re: Spotify’s Discover Weekly: How machine learning finds new music

#32

Post by cisanti »

Asdfbla wrote: post #29

Sometimes I wish Spotify added a bit more 'noise' to their recommendations, so to speak. If I don't listen to much music except Discover Weekly for a few weeks, I (subjectively) find that what's recommended to me more or less sounds the same after a while. Either they are afraid to insert new things that stray too far from an optimal recommendation or they forget too much of my listening history.

I have the same problem. And we it every fad machine learning thing. It never works for me but a lot of people are happy. Does anyone want me as a test subject?!

Re: Spotify’s Discover Weekly: How machine learning finds new music

#34

Post by confounded »

archagon wrote: post #19
aidos wrote: post #12

Discover Weekly works really well for me. It surfaces a lot of music I've forgotten about and for the most part provides an interesting collection to listen to. Having said that, I have a very eclectic taste in music, so it's probably harder to hit on things I won't like. My process is to try to listen to it several times through and then pick out the stand out tracks once everything has had a chance to grow on me. O…

In terms of behavior influence, I listened to pretty much exclusively Hamilton for like a month and didn't get very many new Broadway or hip-hop songs (which is good for me because I tend to not really listen to either genre). So there must be significant weight given to songs/genres you've historically listened to a lot.

Or, based on the way the algorithms work, people that listen to the Hamilton soundtrack a lot, don’t listen to hiphop/broadway a lot.

Re: Spotify’s Discover Weekly: How machine learning finds new music

#35

Post by jordanthoms »

yilugurlu wrote: post #5

They're not anywhere near as last.fm's recommendations.

Last.fm's customized radio didn't work well for me - it'd play the same limited selection of tracks every time, often one song from an album and never play any of the other ones on it.

Re: Spotify’s Discover Weekly: How machine learning finds new music

#37

Post by smartbit »

chewz wrote: post #24

With all that AI excitement Spotify, Google Play Music and Apple Music misserably fail at generating good musical recomendations. At least for me. Spotify Discover Weekly was no different for me when I still used Spotify (six months ago). Not only it is not personalized at all it is also quite lame. I am doing enough to seed - following enough artists, genres, liking album, subscribing to playlists but the quality of…

Same here with Netflix: their recommendations have been a complete failure. It seems like Netflix is pushing mostly their Netflix original series. What does works for me in Netflix are the similar to this below the movie, e.g. after you've viewed the movie. But that is a very simple algorithm any sophomore could write, not related to the recent developments in neural nets/machine learning.

When I go to an filmfestival or an concert, part of what I pay goes to the proffessionals who make the selection for me. And I'm happy to pay their services, just like I pay for journalists.

Re: Spotify’s Discover Weekly: How machine learning finds new music

#38

Post by bbx »

erokar wrote: post #26

A curious side effect of Discover Weekly is that it sometimes influences what I listen to, i.e. I'm afraid to listen to a song in a particular genre because I don't want Disover Weekly to get the wrong idea. But then again, maybe Discover Weekly knows me better than I do.

I had the same issue: I spent a couple of days listening to old tunes I rarely pick anymore, and it “messed up” my Discover Weekly. I thought “Well I only have myself to blame, don’t I?”.

Turns out Spotify has a “Private session” mode to prevent influencing your taste.

Re: Spotify’s Discover Weekly: How machine learning finds new music

#40

Post by w0utert »

tjoff wrote: post #25

Discover Weekly is weird. For the longest time it kept recommending Finish music (roughly 30% of all recommendations), I'm from Sweden and barely ever listen from anything from Finland (or Sweden for that matter). They have a ridiculous bias towards covers. I bet I've been recommended 50 (I wish that was an exaggeration) version of the Gladiator theme ('now we are free'). And they are all terribly bad (as in blood co…

I have a similar problem with remixes from trance tracks I listen to regularly. Some of them have literally 30 different remixes, and they keep getting into the 'discover weekly' playlist. I guess these would be coming from the acoustic profile or the NLP recommendation algorithm described in the article. That said, I still like the feature a lot, it's a little naive to think it could only ever recommend songs you will actually enjoy.
Post reply on HN