Maybe, but by setting it to newest, this is telling me it’s not learning my listening habits very well.
I confirmed that this feature is hosed. When I saw the soundtrack to the Cats movie, I knew it was FUBAR. Cats? Cats! Really??
Yes! Turn off the “spam” and I’ll be happy. The recommendations are nothing I ever listen to, or ever will want to listen to.
I have to say this feature really is dead in the water for me. Sorted by relevance nothings been added since.i cleaned up a good few weeks ago. Go to date and it’s all classical. Really Roon I can’t say I’ve listened to that much classical music in the last year to warrant how much it puts in.
Agree, would like to see a feature to disable it if you don’t like or require it.
Have you been finding a lot of other music you like via traditional means?
It seems like there have been a relative dearth of good new releases. If true, that would explain the absence of decent recommendations.
As for the classical stuff, I also don’t understand how some of these things end up in my bin, as I virtually never listen to classical, although there is some in my collection. But in the absence of good new music, it doesn’t have a good look.
Yeah, but not added much for a while that’s actually new mainly older stuff that I got through the album recommendations which works very well compared to this. It really is a big disappointment.
Found 3 albums just by looking at Qobuz new releases that should have been recommended based on my listening habits and library.
Can you share what those albums are with us so I can pass it over to the team?
Sure I would have expected the following to have been there. The Archer by Alexandra Saviour, Making a New World by Field Music, Live with the No.6 Orchestra by Everything Everything. There are more to Velvet Mood by Billie Holiday Lonely Generation by Echosmith and more. Instead I got nothing new when sorted by relevance and mainly classical music for sorted by latest. This morning a few new ones have crept up but still not all the ones I would expect and I have not added them to my library.
This week Roons mixed things up now with the usual classical music it now appears to think I like Metal🤦♂️. I Sorry is this feature just a joke it has not an idea of what I like at all?
I’m also getting Metal. I don’t thunk I’ve ever played a metal track. Maybe by accident as a radio selection. I’m not sure there is any personalisation at all from roon. Same goes for radio, discovery, etc. Aren’t these selections really aggregations of community taste? Maybe roon should call this feature “New releases for the community”. It would be easier to ignore.
I find Radio works better for finding music to my tastes. This on the other hand is a waste of time and screen space.
I’d guess that is true to a certain extent. There is probably an algorithm of some sort that is tuned by community-wide feedback. For instance, in Roon Radio, the global (community-wide) thumbs up and down of suggested tracks tune the algorithm. The personalization would come from user-specific inputs, for example your library and listening history.
Roon Radio has had a while to get tuned and, importantly, gets lots and lots of feedback for the tuning. I’d suppose the ‘tuning’ analogy for NRFY is what recommended albums get played and what get dismissed. If this is true, dismissing unwanted albums and adding interesting albums would be useful feedback. What would also be really useful are albums/tracks that are missed. Hard to know if Roon has a mechanism to capture this. I suspect this is why direct feedback would be helpful to Roon and is why Roon staff are asking what has been missed that you expect to see.
All of a sudden I got this inspiration:bulb:
It occurs to me that there is no way of telling Roon what kind of music you like.
We know that there are a lot of different opinions out there about deliniating genres and we also know that the genre descriptions in Roon are kind of broad but they should work as boundaries for recommendations and radio.
Wouldn’t it be a good idea to let users specify their likes somewhere in Roon? There are already a gazillion things you can specify/filter/always use/never use/praise/gripe about, so what’s one more?
This one might actually add a lot of value to Roon. Me, I’d specify classical, metal and jazz as explicit likes. That would spare me having to grumble about Céline Dion, Rihanna and all that country music Roon is hell bent on proposing to me even though I never - and I mean never ever up to the point of exhibiting superhuman reflexes to press “next” whenever Roon radio throws up this stuff - listen to that kind of music.
I think something like that should be feasible enough. Just some filter criteria for the algorithms to apply post-query pre-return.
That kind of personalization would really improve the Roon experience IMO. Especially if we could also express explicit dislikes.
That’s an interesting idea @Frank_Daman. We know you can BAN an album, artist, track and it won’t get played by Roon Radio, for example. I wonder if a banned artist would show up under NRFY?
If you could extend a ‘ban’ to something like a ‘not this genre’, you could weed that stuff out.
Roon knows what you’ve played and how often you’ve played it. Problem is it’s relying on lowest common denominator in deciding what to present you, so if the LCD of your listening also listens to rap, that’s what you’re going to get. Personally I think using a community’s listening habits to drive machine learning is poor idea - doubt the population generating the dataset spans enough users to weed out the crap, so you end up with generalised recommendations.
A machine learning engine powered by understanding of musical attributes would be far better than machine learning fed a community’s collective listening habits.
Roon knows what you like by what you have favourited, listened to etc. As far as I can see Ron radio uses the information better than releases because as @evand says releases are generic, radio is more personalised.
Not fully, just more.
What music services are doing this now @evand?
Not sure, I think Gracenote does (what used to be MusicIP’s tech) as does Spotify (remember echonest). Both were vastly superior compared with Roon Radio, albeit they didn’t use machine learning, only acoustic analysis.
The AcousticBrainz initiative could possivly provide the analysis component for Roon to combine with individualised listener data and machine learning.