Hacker News Reader

Comment by Uhhrrr | Hacker News Reader

Rather than subjecting oneself to the whims of recommendation engines and random people, I recommend starting with the artists you like:

-Find out who they've toured with - this is one of the strongest signals possible

-Find out who produced their albums, and check out other albums produced by those folks

-See if people in their band(s) have released their own work

-Read interviews to find out who their influences are and who they like currently

-If they're on a small, focused label (Amphetamine Reptile, Elephant 6), go listen to their labelmates

View on HN

This doesn't get me out of my bubble.

I'm a huge fan of cheesy synthwave and would have never known had Trevor Something - Death Dream not been a freeleech pick on w.cd years ago.

Curation + association is the key.

Not necessarily - a lot of musicians listen to stuff which is not obviously similar. Examples I've seen: Billy Gibbons -> Depeche Mode, Robert Plant -> Patty Griffin, Bono -> Frank Sinatra, Lars Ulrich -> Lou Reed -> Laurie Anderson, Frank Zappa -> L Shankar -> Stefan Grappelli.

But yes, if you really want unexpected paths, you have to just keep your ear to the ground. The way I discovered I liked Goa music was that I randomly clicked on it while surfing Shoutcast 15 years ago.

EDIT: (of course the disadvantage is that you have to sort through a lot of chaff)

w.cd had the best recommendation system for music and audio equipment. The communities that tried to take it's place after don't come close.

It's a shame because I don't know where to talk about it anymore. r doesn't come close, at least not yet. The community music collages on w.cd were unreal.

yeah, we really lost a huge resource when w got shut down. I had mostly stopped pirating music by then, but still visited the site daily just to find new recommendations. r seems like a good place to acquire music, but has completely failed to recapture that community aspect.

The influences does, but what you really need is an engine that recommends bands based on shared influences.

In other words, two bands in totally different genres who both call Led Zeppelin an influence

I get what you're saying but how does that help with small name or indie artists like Blank Banshee or Trevor Something for example?

What you're suggesting is definitely feasible with the larger acts that have been around and have been thoroughly psychoanalyzed.

Usually the bands themselves will tell you their influences. It's a very common interview question, and often appears in bios.

I'm a big Trevor Something fan as well (love that he releases an LP or EP annually), but surely you'd have noticed his material on NewRetroWave, the biggest synthwave Youtube channel.

They love TS as well, and make sure to upload his new stuff with a custom video.

Before that freeleech pick I had no idea any of that universe existed. I had no idea this genre even existed. Now I do.

Producers are a huge piece of the puzzle to unlock. They tend to stick in similar areas of music, so odds are, you will like the other stuff they work on. Great point.

Sounds like every one of our tips could all be automated into a.... recommendation engine. Almost trivially...

And yet all recommendation engines have historically been crap

All good points. But why doesn't Spotify, Apple, etc. either make this node options available to traverse the library? Or at least incorporate it into their algorithms to improve suggestions?

I find Spotify's suggestions closer to "Here's things we're confident you won't hate." Which makes sense if "engagement" is their measurement. But by my measurement, it's yawn-city. Beige.

My best source used to be BBC Radio Music 6. But about a year ago they shuffled their DJs, as well as the programming. Still better than Spotify, but it's rare I'm wow'ed. Previous to the change it was weekly.

p.s. I work from home. Traditonal radio often feels like a convenient proxy for human contact. Just me?

I used to be quite happy with Spotify suggestions, but these have become much worse lately. Mostly remixes of what I already listened to. And the new app ux feels more unintuitive, adds friction (way more clicks). I am at the point where I start looking for alternative music players.

I'd say it's because Spotify and others don't have an incentive to let users dig that deep. They want to keep things simple, whereas sites like Discogs and Rateyourmusic are designed to let you go deep into the details.

Streaming networks need users to be able to see something they can hit Play on, not have the user click on a producer's name and see no information and no tracks.

It's a graph database. Nothing esoteric by any means. I shouldn't have to know Jagger & Richards are The Glimmer Twins, or Richard D. James is the Aphex Twin.

If Streaming Service X's value add is beige recommendations - and the music industry is content with such "product placement", then they both deserve to suffer. If only that pain was 10% of my boredom.

True there was a song from a band I used to listen to and I realized it had a name from someone not in the band and found out they were in another band too. It was slightly different style but interesting and enjoyable. So it pays to look further into artists and people involved.

Can you build me an app for that?

That is maybe a good idea! If I were going to do it, I'd start by crawling AllMusic and select portions of Wikipedia. But it's easy enough to do it manually for the use case, "I've run out of stuff to listen to by X, how do I get more?"

I had attempted to do this four years ago but had hideous experiences with trying to find someone to assist me with execution.

When I was working on it, it seemed a lot simpler to query the Musicbrainz API for a lot of the information.

I'd like at least some graph database containing all these entities. Can probably be scraped from the internet, would need to know which ones are highest quality and most comprehensive though.

what you are looking for is MusicBrainz. You can download their full db and start hacking from there.

Support acts are often mismatched. Lots of acts tour with their mates rather than their musical peers

which is a great way to discover new and different music - the sort of thing the algorithms won't recommend to you.

Algos have been much better to me tbh.

the discovery algorithms are pretty awesome in general, but they're really easy. nobody's wondering how to find the sort of recommendations that spotify gives you, because everybody gets plenty of algorithmic recs every day.

but sometimes it's nice to get something that goes a bit weirder than what the algos typically give you.

Same could be used for movies - directors, cinematographers, producers. I wonder what else this could be applied to?

It works for literature reasonably well. Start with somebody you really like, and see who they wrote critiques and reviews for, who they wrote letters to most frequently, and who they admired or otherwise aspired to be more like.

Works for software, too. What’s new is old, and the old stuff is often quite good.

There’s a lot of stuff that was way “ahead of its time”, to borrow an aphorism, but nobody knows this because it hasn’t been “rediscovered”, to borrow another.

Reply on news.ycombinator.com