introductory thoughts on a playlist generation task in MIREX 2009

I'm currently knee deep in ISMIR activities, of which I will do a more thorough write up when everything has finished. Now, however, I think it will be useful to briefly discuss a MIREX 2009 task I'll be pushing for, that of automatic playlist generation. I think it will be very useful to formalize as much of this task as possible as soon as possible, in effort to encourage participation and avoid the nearly standardized MIREX august rush.

With that in mind, here some the basic starting questions followed by my [brief and biased] thinking on them:

1. What is a playlist?
   
   

   From wikipedia:

   In its most general form, a playlist is simply a list of songs. The term has several specialized meanings in the realms of radio broadcasting and personal computers.
   The term originally came about in the early days of top 40 radio formats when stations would devise (and, eventually, publish) a limited list of songs to be played. ...
   As music storage and playback using personal computers became common, the term playlist was adopted by various media player software programs intended to organize and control music on a PC such as Nexus from NexTune. ...
   Some websites allow categorization, editing, and listening of playlists online, such as Project Playlist, Plurn, imeem and Webjay. ...
   

   
   This is a sensible starting place from a broad definitional stand, especially when we acknowledge that the exact MIREX will obviously be a subset within the idea of a playlist (more on that later...)
   




2. How do we quantitatively evaluate an automatically generated playlist?

   Upon a brief reflection this is clearly the sticky bit. As such this question is being posed prior to any sort attempt to specify an exact task. Functionally speaking, if the evaluation is properly specified the ask description will practically fall out of it.
   There seem to be a few approach to this, though many tend to revolve around an idea of mining some form of ground truth out of crowd wisdom (here and here for instance)




3. And of course, what exactly should the task be?

   Again, this is of course all very draftish, but my prime thought on this is to have a highly specific task of one (or all) of the following forms: query with song A B C algorithm generates song D (and perhaps song E, where order matters), the simple deviation of that form given A B E, generate C D. The other possibility is something alog the lines of give a query song and provide the next song that should occur in the list.

So anyway, those are my initial thoughts on this potential task for next years MIREX. Please comment if you have thoughts...

A brave new world

This morning as I 'opened' my NY Times, what do I discover but a review of the new google browser, not in the tech section or the style section, but as an editorial piece. Who would have thought that a New York Times editorial would read like it was on Ars Technica. Clearly these are interesting times.

in between an ICMC and an ISMIR

So I'm gathering my thoughts on ICMC (I presented this poster) while simultaneously prepping for my poster/presentation for ISMIR (paper).

ICMC was very interesting. Kurt has a nice summary of a few of the notable papers. I did actually go to Bob Strum's presentation on Friday and there tech has some amazing potential as well as a rather slick interface (especially for research software...). Beyond this though, I spent a good portion of the paper sessions attending the discussions on aesthetic and philosophical considerations, in part because I have approximately no formal background in that sort of thing and as I'm setting out to make playlists (and mix them) in a way that's pleasing (or at least interesting). So I heard a very interesting talk by Gary Kendall presenting an attempt at an event based descriptor schema for Electro-Acoustic music (can't find the paper online...). While I must admit that the details of the schema's application to Electro-Acoustic music left something to be desired, I found the general constructs of attempting to define event based descriptors of things like expectation and energy expended to be very intriguing.

I have some video from the mobile phone orchestra performance, that I'll be posting a bit later, providing it's worth looking at.

Back to prepping for ISMIR.

A mashed08 debrief

So over the weekend I attended the hackfest that BBC puts, Mashed08. It was a 2 day marathon sort of hackfest where BBC and other sponsors (Yahoo was pushing their new locator techFire Eagle for instance) walked you through some newish API or otherwise accessible tech then you had 24 hours to make something rad. Each sponsor gave out prizes in the end for the hacks they judged to be the best at using their bits.

Yves, Kurt and I set out to make something rad as well. The basic idea was to recommend BBC radio brands (shows) based on the music in your personal collection. We (mostly, kindof) made this work.

The idea is that you generate rdf describing your audio collection (or a subset) using gnat, which is a piece of the music ontology tools. This rdf file contains the MusicBrainz IDs for your tracks which can then be cross referenced against the available playlists for BBC radio shows in order to find the best matches.

The biggest problem we encountered (which seems obvious now but didn't when we started) is that a collection of non-trivial size generates way too many results to be useful and we didn't implement any sort of ranking scheme to present the results in a way that is useful, though this could easily be done and maybe one of us will add this at some point...

Also, there was a tech crunch write of the weekend that shortlisted our project, which is neat. (The Guardian has a more comprehensive article as well). Finally, for posterity:



more pictures

Needs something to do tomorrow?

I'm going to be DJing at this event at the Dana Centre in London tomorrow. The playlist for my DJ set has been automatically generated using both social network information (from Myspace) and acoustic similarity measures. If you're intrigued, come to the event. if you can't make the event, the algorithm will be presented in an ISMIR paper this year, so you'll be able to see a bit of it there (if you're the ISMIR attending type. You know who you are...)

Greatest example of a need for data compression...

From the duke.
I believe the entropy of that set is actually zero, since both artists are functionally the same, so with an nice compression algorithm with the same two hour fixed transfer time, the transfer rate goes to infinity.

ha!

wherein I poke at thisismyjam

as many readers of this blog probably already know, the echo nest recently announced their new web service based platform analyze and the platform's first app (written by them), thisismyjam.com. I went over there and checked it and while it's not without it's problems (perhaps most notably drastic song to song level jumps and a lack of phrase alignment) it's a fairly decent web app and has me wanting to play with their API, so it has achieved that goal at least.

You want to see my jam? reich + cage x ad nauseum =