Next week I am going to visit the University of Montreal to give a talk at Professor Eck's 'Music and Machine Learning' course.  Doug asked me to suggest a paper or two to serve as background reading.  Instead, I'll describe the prerequisites right here in this blog entry.

I'm going to talk about music discovery and recommendation.  Since the course is about music and machine learning, naturally I'll talk about how we can use some of the automated techniques such as music classification and music similarity to improve recommendations. But before we get there I want to put the machine learning bits into a larger framework of music recommendation and discovery.  To fill out some of this background and to get you thinking about music recommendation,  you have two pre-class assignments:

The first assignment is to visit and use the following music recommendation/discovery sites:

(Update: Note that Pandora might not work in Canada, GlobalPandora might work, but then again it might not)

For each site think about:

  • What technique does the site use to help you discover music?
  • How far into the long tail does the site get you?
  • How deep is the catalog of music?
  • How central is music recommendation and discovery to the site?

The second assignment is to post a comment to this blog entry that describes how you made your most recent music discovery.   I'll post the description of my most recent discovery to kick things off.

(note, if you are not taking the class, but just happen to be reading this blog, you don't have to do the homework, but you can if you want).

Update:  I've posted the slides for the talk. 

 


Comments:

My most recent discovery is Kate Havnevik. I discovered Kate by listening to Pandora. I was listening to Pandora's "Bjork" radio, which plays music that sounds similar to Bjork. Over the course of a few days, Kate's song 'sleepless' came on a few times. After a few listens I decided that I really liked it. I checked how Kate's wikipedia entry - I noticed that a number of her songs are featured in the television show "Grey's Anatomy". My high-school-aged kids really like that show, so Kate was my way to have a conversation about something that interests my kids (kind of sad really ...)

You can hear 'sleepless' in the Youtube fan video:

http://youtube.com/watch?v=sq_B78bKnz4

Posted by Paul on January 22, 2008 at 03:47 PM EST #

My last discovery that is worth talking about is a rap group from Quebec called Gatineau. I went to the "Festival Consonance", a small music festival in St-Hyacinthe, Qc. [http://www.myspace.com/consonancefestival]. I already knew some groups that were there, but it was the first time that I was hearing about Gatineau.

I really enjoyed their show although I've never been much of a rap fan! It's probably the reason why it's the first discovery that came up to my mind, I discovered a new group that made me re-discover a music genre I've never liked before.

To know more about Gatineau:
http://www.gatineau.mu/
http://gatineau-c4.blogspot.com/

Posted by Lucie on January 26, 2008 at 03:57 PM EST #

Most of my music recommendations have come through a weekly podcast from CBC Radio 3. I hear new artists on their weekly podcast every week, so it's hard to pick just one. I think the one that's had the biggest impact is an indie artist from Halifax, Rich Aucoin. His music is great, and as a bonus it syncs up with "The Grinch who Stole Christmas" quite nicely.

Some further links:
http://www.cbcradio3.com (http://bandeapart.fm is the french version)
http://www.myspace.com/richaucoin
http://www.youtube.com/watch?v=f1GVcn_230U (his full album with the Grinch. The individual songs are in the related videos.)

Posted by Andrew Hankinson on January 27, 2008 at 03:29 AM EST #

Quite a few of my recommendations come by seeing the bands that open for artists I’m already a fan of. The most recent example was Au Revoir Simone, a charmingly twee band from Brooklyn who opened for Jens Lekman, an equally charmingly twee singer from Sweden. Sometimes this is more fruitful an area of discovery than others, I suspect this has something to do with whether the headlining artist or the promoter organizes the opening acts.

Posted by Johanna Devaney on January 27, 2008 at 10:17 PM EST #

The only music recommendation site I've used enough to have an informed opinion about would have to be Pandora. I have mostly used it when I just needed some casual background music to work. One weakness of Pandora is that the catalogue isn't very big so you would sometimes get the same songs after a while. I did discover some nice music but none of which I became a big fan of.

I have a couple of friends who have more time than I to look around for new music. Luckily for me, they have a pretty good idea of what kind of music I like so they often recommend me new bands. Many of those end up on my MP3 player playing on repeat so for me, they are the best recommender I've got right now.

I'd also have to agree with Johanna about the opening bands. My latest discovery was an artist called Jason Bajada (http://www.myspace.com/jasonbajada) who was opening for Dumas (http://www.dumasmusique.com), both from Montreal.

Posted by François Maillet on January 28, 2008 at 01:17 AM EST #

[Trackback] Bookmarked your post over at Blog Bookmarker.com!

Posted by grinch on January 28, 2008 at 06:54 AM EST #

My most recent interest in music is ambient electronic labels - many of which now offer free online compilations. The best of these that I have found is:

http://www.sutemos.net

Sutemos is a small Lithuanian label that puts out deeply personal intelligent music that is both progressive in its use of new forms of synthesis/sampling and production, and at the same time classic as it reminds me of lush synthesized soundscapes of early FSOL.

I found this label from a friend of mine, Praveen Sharma (http://www.percussionlab.com), who has been on their last two compilations, and whose music is absolutely brilliant. I met Praveen in NY at grad school, and have used his site as a starting point into a subgenre that I follow fairly consistently.

Posted by Jason Hockman on January 28, 2008 at 08:36 AM EST #

I think I'm too much of a casual listener (ie passive) to discover things through a recommendation system. Music is often an enjoyable backgroung thing, unless it's a piece I already know and like. This means I'm never giving feedback to the recommender, apart from some skips. I'm also not a frequent enough user to get to hear the same new song a few times and start being conciously curious about it.

So my (few) recent discoveries happened through the good ol' buddy system. Here's a friend's band I recently discovered:
http://www.myspace.com/wearewhywhynot

And there's always apple commercials!

Posted by Anonymous Coward (P) on January 28, 2008 at 09:43 AM EST #

Thumbs up for a great presentation! It was a complete overview with concepts seamlessly introduced. I especially enjoyed seeing how I fit in the different kinds of users and what a good recommendation might be (with the roles of familiarity, novelty, willingness to invest in discovery and so on). Thank you!

Posted by Anonymous Coward (P) on January 28, 2008 at 03:30 PM EST #

Post a Comment:
Comments are closed for this entry.

This blog copyright 2010 by plamere