Slashdot Mirror
Going Head To Head With Genius On Playlists
brownerthanu writes "Engineers at the University of California, San Diego are developing a system to include an ignored sector of music, dubbed the 'long tail,' in music recommendations. It's well known that radio suffers from a popularity bias, where the most popular songs receive an inordinate amount of exposure. In Apple's music recommender system, iTunes' Genius, this bias is magnified. An underground artist will never be recommended in a playlist due to insufficient data. It's an artifact of the popular collaborative filtering recommender algorithm, which Genius is based on. In order to establish a more holistic model of the music world, Luke Barrington and researchers at the Computer Audition Laboratory have created a machine learning system which classifies songs in an automated, Pandora-like, fashion. Instead of using humans to explicitly categorize individual songs, they capture the wisdom of the crowds via a Facebook game, Herd It, and use the data to train statistical models. The machine can then 'listen to,' describe and recommend any song, popular or not. As more people play the game, the machines get smarter. Their experiments show that automatic recommendations work at least as well as Genius for recommending undiscovered music."
Stop using the imbecilic term "wisdom of the crowds": aye or nay?
Best thing for them to do would be to crawl last.fm. Some pretty esotheric musical tastes there.
Does this mean more or less Miley Cyrus?
There, now we got that out of the way (and I do feel oddly better about life) I have to say that I'm still skeptical about these algorithms for music recommendation.
A friend of mine and I listen to a lot of the same music. He got me on the soul train in a way, so we talk a lot about Soul, R&B (The old fashioned kind), Funk, Rare Groove, Jazz and Gospel. Now he and I can dig the same song for wildly different reasons. We can sit and discuss the same tune, which we both like, and discover we look at the thing so differently it's as though we're from different planets.
Now I've tried all possible music sites and playlist generators, but at the end of the day I simply never really agree with the correlation they see between song one and song two. I really wonder if the /. audience believes something as complex as music appreciation can be captured in a program....
I'm not a hippy like you, so yes, I for one welcome our song analyzing overlords
So, not really so much at all...?
will be killing us to a bitchin indie playlist?
Science advances one funeral at a time- Max Planck
I know you say you've tried all possible music sites... but on Pandora if you create a new station from an artist or song, they'll give you the criteria they use to populate the playlist.
Set up stations based on enough songs, and it's pretty easy to understand at least part of how their algorithm works. A big problem, of course, is that some of the criteria are somewhat subjective, which is why you may disagree with them. I find this especially true when creating stations based on artists, not songs.
I just wish I could tweak the individual conditions to see where it'd get me... like having all criteria match except genre.
"Trolls they were, but filled with the evil will of their master: a fell race..." -- J.R.R. Tolkien on Olog-hai
The article links to apps.facebook.com/herd-it/?refcode=slashdot
So I'm thinking this is payed advertisement disguised as an article. That's just low.
GAAH! MY PRINTER IS ON FIRE!!! PUT IT OUT! PUT IT OUT!
If there's a reason to the music I like, I would like the computers to tell me. I like all sorts of music, from acoustic folk to pop to alternative rock to christian rock to screamo. I'll even listen to some country now and again. If a music recommender can understand that by my admission to enjoying 38th Parallel and Blindside, that I'd also enjoy something by Jack Johnson, I'd be amazed.
When are people going to realize that unpopular music is unpopular for a reason. Sure the music execs try to push their own artists more than others, and they try to target the largest cross section of the population as possible, but why wouldn't they?
Trying to bring 'unpopular' must to the masses because that will suddenly make it popular is stupid. Music becomes popular because someone hears it and likes it, not just because they hear it.
Throwing Timmy's garage band onto every radio station in the world during prime time isn't going to change the fact that Timmy's garage band sucks and very few people want to hear it.
Yes, there are people who don't have the same tastes as the general public, that is a small portion of the public, nothing you do is going to change that. There will always be a bell curve. Stop with this crap of think just because you like some indie band that no one has heard of that everyone else will.
If the general public likes them they will become popular. If they play a local show and people like to hear them, they'll get requested and more people will hear them. Then more places will request them, and rinse, repeat, until they will become popular.
Unknown bands are unknown because they are interesting or 'good' to a small number of people, not because of some silly idea that they got shafted by a playlist generator. The playlist generator is simply following trends that it learns from people. It doesn't actually analyze the music to find the algorithm that makes it 'good music'. It says 'People that listen to this song also like this song, add it to the list', rinse, repeat, playlist generated. It doesn't say 'hey, no one listens to this song, lets throw it in and then everyone will like it!!!
Persistent Volume manager for Kubernetes - https://github.com/dwimsey/openshift-pvmanager
I tried Genius for awhile, but I recently disabled it for two reasons:
1) The "recommendations" were not very good nor did they maintain a "common theme", by which I mean, I chose a rather edgy electronica/punk song by Crystal Castles... three songs down we get something by The Nationals... which is very mellow rock. If I choose a song that is edgy, electric, and with a faster pace, I want ALL the songs in that 25 song playlist to be at least within a similar genre.
2) It takes up too much time when importing vast libraries to new machines. I recently centralized my 300+ GB music library on a Mac Mini Server, iTunes was unusable due to genius choking on the sheer volume of data it had to deal with.
In the end, it's really nothing more than a way for Apple to try to get you to buy more crap from the Apple store.
They lost my wallet years ago to Amazon MP3 store who had no DRM. I see no reason to go back to iTMS even now that their DRM is gone. Especially seeing what dicks Apple has been with their conduct around ACC, "fair play", and App Store lock-down.
More precisely, popular causes good. Norms cause people to want to act the same way. Some people will listen to music because of its artistic appeal and others will listen to a specific type of music to distinguish themselves from the norm in some way. But the crowd will want to listen to what the crowd listens to *because* that's what the crowd is listening to. Nobody wants to take from the long tail exactly because there's nobody paying attention to the long tail.
I really enjoy classical guitar music. Apple's genius selection for the type of music I enjoy is so bad I just turned it off.
The first track that played was a System of a Down tune. Which is about as pop as it gets. The ones after didn't get much better. If they really want to use this to push less played songs which have potential, they should actually better get some.
The problem with slashdot is that most of its users were bullied and stuffed into lockers as kids!
Capitalism: Hello there Academic. How are you?
Academic: Hi... what they heck are you? You look so strange to me...
Capitalism: I'm Capitalism. Oh, I'm really not all that strange, but I might be a little complicated to understand.
Academic: Complicated!?! I am the master of complicated, I am an Academic for crying out loud.
Capitalism: Ok then. Let me try to explain myself. I am a system that provides stuff via supply and demand.
Academic: Nonsense! I dont hear music that deserves to be heard on the radio or on popular websites!
Capitalism: Deciding who deserves what really isnt my thing... see... its about supply and demand...
Academic: But who decides whats in demand!?! Certainly it cannot be the uneducated "masses", they... just aren't qualified!
Capitalism: No no... its about what many individuals, smart or otherwise, want based on need or dozens of different other factors.
Academic: Preposterous! How could they possibly know what they want or need if they havent been exposed to it?!? Foolish Capitalism!
Capitalism: Well, there are a lot of musicians out there and only so many different ways to get them heard, and, well, there are people out there who spend their lives learning what people like and dislike, and even they arent always right... so the best at determining who does best succeeds...
Academic: Rubbish! What we really need, is for the qualified, with a broad base of tastes to make an application for people to give them a view of all the music that is out there!
Capitalism: I guess you can try, no one can stop you, but you might not succeed.
Academic: Your so short sighted. I don't need to worry about succeeding, I receive public money to pursue my higher realm of thinking.
Capitalism: Right on... so I guess you will compete and regardless if your product sucks, you dont have to worry about it because your really just spending someone elses money.
Academic: Its progress my dear boy. Progress.
20th century Marxism is not progress...
I have to say that the best AI for this kind of thing was Yahoo's Launchcast service (recently sold to, and dismantled by CBS). Almost every day, I discovered new music that I'd never even heard of, and the vast majority actually suited my tastes. Unfortunately, when CBS bought Launchcast from Yahoo, they took out the only valuable part of the service (the "create your own station" part that had the fantastic AI), and they left just plain old streaming radio stations.
I don't respond to AC's.
I've had great success with Gnoosic
The recommendation sites are far from perfect, but they're the best thing out there for discovering new music that you'll actually like (and orders of magnitude better then broadcast radio). Almost all of the CDs I've bought over the past year were of bands I first heard on last.fm. At the very least, last.fm has never played me Nickelback, so its gold in my book.
It's exactly algorithms like the one used by Pandora that make me agree with the viewpoint that it's not possibly to calculate what "other music" I like based upon the "known music" that I like.
Anyone with a preference for Electro Pop will likely have been wondering when the hell Pandora would learn the difference between Miss Kittin and Scooter after mindlessly clicking "Dislike" on eurodance tracks when Pandora fails to see the difference between one type of electronic music with a repetitive beat and another.
The only really worthful algorithm we'll ever manage to produce is one that uses the collective intelligence of all its users.
Stop being arithmetic supergeeks wanting to put everything inside a box, and start figuring out how to get all these weird unpredictable people to input valuable data into your system.
Google figured this out more than a decade ago, so why are we still seeing stupid mathematical and "pattern-based" algorithms every year?
I wrote something earlier this week to do the same thing with the hashtags in the Twitter API and my music DB.
If you're bored, check it out. The recommendations are pretty close (bottom left). Metallica or Weezer
I found this one interesting Beatles because it finds the singers names.
For possibly the great majority of the population, music can be compared to fashion; does not really matter if the art is good per se, what matters is the trend and popularity, on a local scale (what my friends listen to) and global scale (media).
With the rock'n'roll revolution in the fifties, lots of teenagers liked that new music in part because it wasn't their parents' music. Same story can be said of disco, rap and grunge.
Problem with the long-tail approach is that people mostly judge music by non-musical criteria.
Last.fm's "neighbor" system works similarly, except it looks at what each person listens to. Keep in mind that it takes a fair bit of training to find neighbors who are actually close to your likes, but once you've listened to enough music, it's pretty good at finding things I like but have never heard of. I.E. if I like song A B C and D, and you like song A, B and C, you might like song D.
The neighbor system groups people with similar musical tastes, and allows each person to tune to his/her "Neighbor Radio", to listen to songs liked by your neighbors.
(Disclaimer: I have no vested interest in last.fm besides being a paid member. [My Profile])
I'm still skeptical about these algorithms for music recommendation.
You should be skeptical, but not dismissive. In its current state, this type of service is more like directed browsing than a true recommendation. But it still yields the an occasional gem, and with continued participation and increased competition it will get better. Skepticism makes it a useful tool, if you can live with having to wade through some misses along with the hits. Blind acceptance will, of course, be mercilessly exploited and the unwashed masses will still end up listening to the likes of [insert music you despise].
"the most popular songs receive an inordinate amount of exposure"
More to the point, I think the popular songs get played in disproportion to (above & beyond) their popularity. Versus the songs getting inordinate exposure? Anything played on air.
I just wish I could tweak the individual conditions to see where it'd get me... like having all criteria match except genre.
Even a line-item veto for the "why did you play this song" would be ideal. Thumbs up or thumbs down on every song seems to make the music selections worse, not better. Obviously which songs I like and which ones I don't doesn't neatly boil down to criteria that pandora can identify, so I think it's unavoidable that it will pick up on what it thinks I like but I don't.
For example, I like hip hop with clever lyrics, but hate rappers who can only talk about themselves. Most hip hop artists though rap about themselves at least a little, even the ones I enjoy. Simply because all rappers rap about themselves at least a little, any song I give a thumbs up, it's always going to think I like that and will start playing songs that are -only- about the rapper. If I make a station around Aesop Rock, who rap about things besides how well they rap, the selections are good. If I hit "I like" for too many songs though, the selections get worse, looking in "why did you play this song" there is always a line about "boastful lyrics." Being able to select "no, not that one" would be a plus, as the other criteria get better.
(Please, try to resist the temptation to post on how you don't like hip hop. I know, I know, rap music is missing a c, that's very clever.)
hmmm, I checked this out and it recommended a band that wasn't even close to the three bands I gave it. Plus, I didn't like the band anyway. Not promising
I've found a lot of songs/bands I had never heard of thanks to Pandora. I started a station based on "Panic Attack" by Dream Theater, and it's interesting to look at "why was this song selected" for new songs. The current song I'm listening to says "we're playing this track because it features a subtle use of paired vocal harmony, varying tempo and time signatures, chromatic harmonic structure and demanding instrumental part writing." I could have said that I like varying tempo and time signatures, and demanding instrumental parts, but it's neat that it can pick up on things like chromatic harmonic structure and paired vocal harmony.
if your product sucks, you dont have to worry about it because your really just spending someone elses money.
Pretty much the story of most equity-funded businesses (particularly venture-funded.
And C-level agents.
Tweet, tweet.
The neighbor playlist on last.fm is a really effective tool for finding stuff I didn't know about but like.
Is to get 10 recommendations I have to listen to 10 songs in a row due to their skip limitations.
They can keep their skip limitation, that's fine, I get the licensing problem they have.
But why can't I get a simple list of the next N (10, 20...100?) songs they'd recommend based on my current "station"? It might even improve their recommendation engine for me because I could thumbs up/down (and I suppose, "I'm tired of this one", too) the songs and cut through the cruft faster. Sort of like Netflix "Rate Movies" engine which allows for more inputs to the rating system.
I also am annoyed that you can add a *song*, which is highly specific, or an artist which is somehwhat specific, but not highly so given some artists creative changes over time, but you cannot add by *album*.
For example, IMNSHO REM's Chronic Town, Murmur & Reckoning are brilliant and most everything else after is less (often much less) so. Adding those *albums* defines a general yet atomic set of musical criteria; adding the entire band and all its albums biases it much more towards the later records, even by sheer weight of number of songs.
This is true for dozens of bands whose sound changed substantially over time or whose last N albums were horseshit or whatever.
All in all, I *want* to like it but the "service" doesn't mesh with how I listen to music from a practicality sense (streaming via iPhone blows on my car stereo and is a battery hog elsewhere) and 10 recommendations aren't worth 30 minutes of sitting down.
I really wonder if the /. audience believes something as complex as music appreciation can be captured in a program....
It's not like music can be represented mathematically or anything. A computer would never understand that.
I don't know how last.fm works but I downloaded the android application on my phone and so far it is treating me well. It is much faster than pandora and allows me to listen either by entering an artist, tag, or user. So in tags if I type in Jpop I get a bunch of japanese pop songs. I can find stuff from other countries as well which is cool.
It almost feels like iPods are overrated now. It would be cool if I just subscribed to a service and used my phone to stream in music based on my preferences or playlist. Then all music available on the internet (world) would be accessible as long as I had a 3G signal.
An underground artist will never be recommended in a playlist due to insufficient data. It's an artifact of the popular collaborative filtering recommender algorithm, which Genius is based on.
Their experiments show that automatic recommendations work at least as well as Genius for recommending undiscovered music.
At least as well as never recommending? That is astounding.
Let's try an example. Kings of Leon is enjoying some pretty good success right now with songs from their most recent album reaching the top 5 of Billbaord, Hot 100, and other charts. But the band formed in 1999. For for the better part of a decade they were only "good" to a small number as you put it. But then how did they suddenly jump out of relative obscurity to the top of mainstream charts?
You seem to assume that Kings of Leon have produced music for 10 years of a constant quality. Perhaps their "Only By The Night" album (2008) and the associated singles just were so good that the quality of the music made that their popularity grew?
I am not saying that this true, but just because they make good music and are popular now does not necessarily mean that they "deserved" to be popular all along.
Firehed - Unfortunately, thanks to medical breakthroughs, common sense is not as common as it once was.
10 minutes with someone who works in an independent record store will help you find better music than any algorithm (or any Top 40/Adult Contemporary radio station for that matter). That is, if you can find someone friendly in one of these stores...
Paraphrased from actual conversation in an indie record store:
Record store owner: "Why do people keep coming back here?"
Employee: "Well, it's not friendliness..."
Owner: "WHAT?! I'm the most f***ing friendly guy there is!
Freedom is drinking a beer in the park when you're supposed to be at work.
I remember reading somewhere in the pandora algorithm that you should only click "Thumbs Up" for songs you really like, not every song you like. If you like a song, but there are aspects of it you don't like then don't select anything, just let it play.
I know this isn't as nice as being able to select individual features of a song, but what are you going to do?
Sewage Treatment Facilities - "Our duty is clear."
Wow, Grisman would *have* to be a genius to play the mandoline musically, as that's a kitchen gizmo for slicing things. But then I have heard him play the (no "e") mandolin, and yes, the man is a genius, or at the very least extremely much more talented than I am. :) Heck, I think there's even an established style of mando playing named after him -- sure enough, his Wiki article makes mention of "Dawg Music".
Whatever he's playing though, it'll probably sound pretty damn good.
Cheers,
"What in the name of Fats Waller is that?"
"A four-foot prune."
I disabled Genius and just leave it on Random play. I have 35,654 songs, so I basically listen to the best radio station and rarely hear repeated.
Heck, if I live to be 90, I'll probably only hear each song 7 or 8 more times anyway...
Shoes for Industry. Shoes for the Dead.
There are four in my household. I'll use itunes to make up a CD for a long car journey with the kids or a playlist for my wife to work out to. Then there's the music I like to listen to. I'm sure it believes I'm schizophrenic and if it were a real genius it would be giving me psychiatric advice by now.
Nullius in verba
I've played with Pandora radio a bit before, but it doesn't tend to work for my taste in music. I like songs with funny lyrics... Monty Python, King Missile, Nellie McKay... the algorithms that focus strictly on musical styles generate pretty hilarious results, but not what I'm looking for :P
I've been pretty happy with some of the dj internet radio stations, though, like somafm.com and some of the ones on di.fm . Before I found some of those stations, I didn't really think there was much of any music that I liked. Certainly not on broadcast or satellite radio.
They're using people who play Facebook games to train their systems to be smarter?
The mind wobbles.
I know this isn't as nice as being able to select individual features of a song, but what are you going to do?
Whine on slashdot about a free service that lacks one feature I think would be good.
Oh, that was probably rhetorical...
No, I have heard that, but in the case I said, nearly every hip hop artist can accurately be described as boastful, because they all do. Giving -any- thumbs up seems to strengthen that.
But as someone who has spent a MASSIVE amount of time finding and discovering new music I consider community filters like Last.fm to be another useful tool in my toolbox. In fact I wrote a woefully under utilized search engine that combined emusic or amie streets catalog (where I usually buy my drm-free music) with the last.fm api.
:-)
As someone who listens to a somewhat eclectic assortment of dubstep/industrial/electronica with smatterings of indie/electronic pop I do find new music from the suggestions (I'm probably my own best user). I can't speak to pandora because I never really started using them, but if you can't find new music using last.fm you're probably losing patience with it.
Naturally if you end up creating something that provides even better artist recommendations I'll be glad to use it too.
Quack, quack.
My taste in music has been described by more than one person as "random," so yeah, I'm with you.
Pandora has people that hit checkboxes for the songs. It doesn't "pick up" on anything.
If they ask you the question, "What color is this song?", and you answer "Green", but most of the other players answer "Black"..then they ask you that same question again...what are you going to answer THIS time? If they had more than 20 questions to cycle through, they might get useful data. As it is, I think it is just a cleverly designed advertisement for Frank Zappa.
I meant that it picks up that I like that, not that I thought it analyzed the song and determined that.
Blind acceptance will, of course, be mercilessly exploited and the unwashed masses will still end up listening to the likes of [insert music you despise].
I closed my eyes and inserted Nickelback. Seriously, does anyone actually listen to that shit based on any intrinsic merits of the music? Or is it just because it's what the radio plays and who gives a shit that every song sounds exactly the same except for random variations in the lyrics? Hell, in 10 million years Nickleback might manage to evolve and actually create a new piece of music. Darwin would be so proud
Celebrity worship is a poor substitute for Deity worship and costs more to boot.
They invent an algorithm that will always recommend old Metallica over new Metallica!
Just wait 'til they get these things recommending foreign music. I can't say whether that's good or bad, really. On the one hand, I don't want to hear any more Daler Mehndi than I have to... but on the other hand, I would've liked to have been recommended Coeur de Pirate a long time ago.
I just wish I could tweak the individual conditions to see where it'd get me... like having all criteria match except genre.
This seems to be the point of this Herd It game: they've released a beta search engine that lets you find music using words - like genres, instruments, emotions etc. It seems to build it's results from the collective intelligence of all the people who play the Herd It game as well as music analyzing algorithms. You have to sign up for an invite on the Herd It game page.
It doesn't have a lot of words yet (maybe 100 or more). I searched for "funk", "horn section" and "at a party" and got James Brown and Rick James - not bad. But "romantic jazz saxophone" gave Chet Baker and Miles Davis - seems like it can't tell a sax from a trumpet.
Please read the GP. He's talking about the difference between Miss Kittin and Scooter. Scooter! There's a difference between "broadening your taste" and licking out a septic tank.
Isn't that how we keep winding up having to choose between two douchebags for President every 4 years?
I'd like to see a test of Zune's Smart DJ. It seems like it works wonderfully. Unlike Genius it can handle music the Marketplace knows nothing about. I downloaded a fresh J-Pop release that the servers could have no information on, ran Smart DJ on it, and got a playlist full of music that matched not only genre but also generally fit the beat and tempo of the originally selected song. It seems more Pandora-like than Amazon-Recommendation-like. Smart DJ also doesn't need to do any long analysis of your library.
And then what happens when all the music in the world is 'discovered?'. There will be an uprising like seen no other! There has to be undiscovered music to keep the malcontents at bay.
Haha well said.
Back in the days of Napster and others of it's ilk, I could log in, select a random sampling of tracks from bands I'd never heard of, listen to them, and then go hunt down the band if I liked what I heard.
About 80% of my music listening, other than classical, was discovered that way. I might have 10-15 "established" bands in my playlist now. The rest are all indie or self-published.
Sadly, there's still no "legal" way to do that kind of random sampling anymore.
Napster, Grokster, mp3.com, others... have all come and gone.
Now I'm stuck with using ... well. never mind what I'm using now. I'm not going to violate rule #1.
Because once this shit gets 4chan'd, the shit hits the fan and you got crap data.
Nice conversation. If anyone having problems with their mac then use this free tool Mac Data Recovery.
Jolie Joe http://www.datarecoverymac.org