Wednesday Aug 29, 2007

If you have a large music collection you probably know all about messy metadata.  Artist, song and album name misspellings are common.  Missing or incomplete data are par for the course.  Inconsistent numbering and formatting, improper internationalization, duplicates, partial albums, multiple encodings, compilations - all make this a very sticky problem.  Now, imagine you are a music researcher with 100,000 tracks, or even imagine you are with millions of tracks, all with messy metadata.  The messy metadata gets in the way all of the time - it ruins recommendations and playlists, it confuses, and just makes you look clueless when you can't tell that ELP, Emerson, Lake and Palmer, Emerson, Lake & Palmer, EL&P are all all the same band. has decided to take on this problem and solve it. But not just for themselves, but for the world. They are distributing an audio fingerprinter that will collect data on common misspellings for tracks, artists and albums. Soon they will be offering web services that you can use to clean up the data.  This will be a boon to all of mankind.  RJ describes the project on the blog: Audio Fingerprinting for Clean Metadata

 I really hope this will tie in with the MusicBrainz database.

Update:  Elias points to this post on the MusicBrainz forum where Russ (of clears things up:

We have no intention of dropping MusicBrainz (especially since it's 
taken so long to get the license in the first place!). MB is a lot more
than just the fingerprinting: we think MB is a great source of metadata
- and the more metadata sources we have, the better. Our fingerprinting
services will definitely return MBIDs in the future.

That's just super! Now hopefully 'the future' is 'real soon'.

Tuesday Aug 28, 2007

Roger Dannenberg  (trumpeter, composer, computer scientist) forwards some images of his music lab at Carnegie-Mellon University.

Roger's lab is located in Wean Hall, the main home of the Carnegie Mellon Computer Science Department.  Roger says "the banana trees are a nice touch -- it was about 90 degrees today, and I think this is some kind of joke by the buildings and grounds people. They'll take the plants out in September. Obviously not a native species."

Here's Roger (altthough he says he usually works at a laptop elsewhere in the room).

Hyun Kim is working with Umpei Kurokawa over the summer on audio-to-midi
alignment features for Audacity.

Umut Semsekli is visiting from Turkey and working on accurate beat
estimation from foot-tapping as a component of my work on popular music

Roger says that he would like to have included Ning Hu, finishing her Ph.D. thesis while working at Google, and Wei You, finishing his Masters thesis while staying with family in Philadelphia.

Wednesday Aug 22, 2007

Idiomag gives you your "own personalised music magazine".  Every day, Idiomag will select a few articles related to bands that you are listening to and present them with a  very slick flash interface that resembles an interactive magazine that includes images, music, video and text.  The content is not your typical music blog stuff (which, quite frankly can be of questionable quality), Idiomag licenses professional content a variety of sources including Losing Today, Sounds XP, and Music Emissions.


How does Idiomag know what music you like?  You just point it at your, MyStrands, iLike, or MOG account.  They apparently will suck up your taste data (that sucking sound is the sound of terms-of-service licenses being violated) and use it to figure out your musical tastes.  If you don't have one of those accounts,  no worries, you can tell it a few bands that you like instead.

 Idiomag will make their money by selling ads targeted at their listeners. (I didn't see any ads during my tour of the site). 

 I like the slick interface of Idiomag - they've done a very nice job with that ... but I can't exactly figure out they picked the artists for my magazine: TV on the Radio,  Lavender Diamond and Machine Head.  I haven't listened to these bands in the last week (as far as I can remember).  Idiomag doesn't tell you how they selected them (in fact they don't tell you very much at all - they don't even have a privacy or terms of use statement on the site as far as I can tell).   I am extremely curious if they've licensed the taste data from iLike, mystrands, or MOG.  Idiomag's usage of this data is clearly commercial - and is probably violating all sorts of terms-of-use unless they've licensed the data.  But I guess that's the web 2.0 way of things ... build your site, take what data you need and worry about the terms after you've built a user base. Thanks to Oscar for the tip!

Dog Cove
Originally uploaded by squamloon.
It's a vacation day today .. I'm off to Squam lake for some canoing and kayaking with my dad and my son. It will be my last outing with Chris before he heads off to college on Saturday. It looks to be a beautiful day for it.

Tuesday Aug 21, 2007

We are building a  system that is trained to predict social tags for music.  Our system will try to 'learn' what social tags should be applied to an mp3 file based solely on the audio content.  The system works rather well for most kinds of music, but we've been seeing some particular problems with music that is labeled 'grunge'.  Bands that you'd think would be labeled as 'strong grunge' - bands like Nirvana and Pearl Jam - are only mildly grunge according to our system.  This has been a bit of a mystery, until this week when we added a feature that lets us search through our predicted tag space - with this feature we can find the most 'metal' (Good Riddance, Atheist), the most 'folk' (Bob Dylan, Cat Power), and the most 'electronica' (aphex twin, Jake Fairley).   However for the most 'grunge' we didn't see Nirvana, or Pearl Jam (nor even Courtney Love), instead the most 'grunge' artists were Johann Sebastian Bach, Charlie Parker, Glenn Gould and Wolfgang Amadeus Mozart.  Clearly this isn't right - so we dug in to see what was happening.  We looked at the training data for the 'grunge' tag - sure enough it was filled with non-grunge artists, classical, jazz and spoken word - so instead of learning 'grunge' we were learning randomness (and woe).  Thierry quickly found and fixed the problem (it was a problem with how we were normalizing tag counts) - so now we are retraining a bunch of our tags - with hopefully even better results coming.  Clearly though, we need to think a bit more about adding some basic sanity checks - so we don't happily go though life thinking Bach is grunge.

Friday Aug 17, 2007

One of the simplest ways to publish a playlist is in m3u format.  This is just a list of files or URLS to mp3 files.  Almost all music players support them.  However, iTunes makes a botch of it.  If you play an m3u file in iTunes, it loads the links to the songs, and then just tosses them into the main library list, plays the first song and then then plays the next song in your library, instead of the next song in the m3u playlist.  In iTunes, M3Us become unusable.   To deal with this, there's M3U2Itunes, a freeware helper application that fixes these problems.  M3U2iTunes simply creates a single playlist named "M3U Internet Stream" and places all songs from the downloaded m3u playlist into the M3U Internet Stream playlist, and plays the playlist.  It cleans up nicely after itself too, so you don't end up with hundreds of playlists in iTunes.  The program is easy to install, with good instructions showing how to make finder and Safari use it instead of the default iTunes behavior.   M3U2iTunes is a good app to have.

Thursday Aug 16, 2007

An oldie but goodie from the onion:

Area resident Pamela Meyers was delighted to receive yet another thoughtful CD recommendation from Friday, confirming that the online retail giant has a more thorough, individualized, and nuanced understanding of Meyers' taste than the man who occasionally claims to love her, husband Dean Meyers."I don't know how Amazon picked up on my growing interest in world music so quickly, but I absolutely love this traditional Celtic CD," Meyers said. "I like it so much more than that Keith Urban thing Dean got me. I'm really not sure what made him think I like country music."

Tuesday Aug 14, 2007

I did quite a bit of coding over the weekend while off the 'net - which is a bit like walking the tightrope without a net, since I can't check in my changes into the code repository.  This morning my app just stopped working  - bad news - without any CVS history to save me.  Ah, but then I remembered netBeans 6 has a new local history feature that lets you see your changes and revert to previous 'versions' of the code, even if you haven't checked the code in yet.   With this feature, I was quickly able to find the place that I accidentally munged the web.xml file while trying to add some Security Constraints.  I reverted to the working version and Voila! - I was back in business. Thanks Netbeans!


Monday Aug 13, 2007

There have been a few stories and blog posts floating around that suggest that Universal is partnering with Google to sell DRM-Free music.  These stories are just plain wrong. These stories started from a Universal press release that listed Google as a participating vendor for its new DRM-free music offerings.  This was an error - according to Forbes, what Universal really meant was that Universal would purchase Google ads to steer customers to music vendor gBox.   gBox has nothing to do with Google (which is instantly obvious as soon as you see the 'Sorry we only support Internet Explorer' page at the gBox site).   Universal or gBox buys a Google ad for Amy Winehouse and directs the click-throughs to gBox where gBox sells a track.  Universal and gBox both get a cut of the sale. Google gets the ad revenue. That is not a Google music store.

Google reaffirmed that it has no plans to open its own online music store.

Wednesday Aug 08, 2007


A few months ago, Amazon and Tivo rolled out their Unbox on Tivo. With AuoT, I can rent a movie from Amazon, have it downloaded to my TiVo  so I can watch it on my TV.  Amazon gave me $15 credit to try it out, so I finally gave it a go this week.    I really like the idea of AuoT - select a movie from a huge collection of movies, pay a reasonable price and watch it immediately.  Unlike a rental store, I don't have to make a special trip, nor do I have to worry about late fees, forgetting to put the disk back in the box, scratched/unplayable disks. And unlike Netflix, I get more immediate gratification - I don't have to try to guess what I am going to want to watch in a few days (I'm embarrassed to admit how long I had a Netflix envelope with Seven Samurai sitting on the TV waiting for me to be in the mood to watch it).  Overall, the AuoT experience was Okay - I like the convenience but there are some problems:

  • Selection:  there are less than 2,000 movies to chose from (compared to over 80,000 at Netflix)
  •  Price:  $3.99 is higher than I pay for most movies at the video store - while the cost for Amazon to deliver the movie must be so much less than Netflix or the local VideoMat
  • Play Restrictions: Once a movie is downloaded to my TiVo, I have 2 days in which to play it - and once I start watching it I have 24 hours to finish it before it is deleted from my TiVo.  That's just stupid. The rental period should be longer (3 days at least), with no restrictions or forced deletions in that period of time.
  • Video Quality:  The picture for the most part was very good - better than broadcast, but not as good as DVD. There were some stuttering artifacts every once in a while.
  • Time to Download: It seems to take about 2 hours to download a 2 hour long movie to my TV.  The annoying thing is that you can't start watching a movie until it is completely downloaded.
  • No Recommendations:  Uncharacteristically, Amazon offered no recommendations as to what movie I should watch - they seemed to just list movies in order of popularity.
I hope Amazon/Tivo adjust things a bit to make this a better experience.  If they lower the price, increase the catalog size, and eliminate the stupid play restrictions, I'll visit them more often that the local Video store.

Monday Aug 06, 2007

This month I'm featuring photos of the various music information retrieval labs that are the brains behind music 2.0.  

This week features  CIRMMT - the Center for Interdisciplinary research in Music Media and Technology located at McGill University in Montreal Canada.

Ich has sent me this nice panoramic photo of the lab.  Click on it to see the full size (and very large!) version.  There are lots of nice things to see in this shot.  Everyone seems to have nice dual flat-screen displays, a mix of Apple and PC workstations, lots of natural light, people working together (except for poor Cory, all by himself in the corner).  It looks like a fun and productive space. 

This picture is of the working space - on my tour of CIRMMT last year, Ich showed me a whole lot more -  the CIRMMT web page describes the various CIRMMT labs:


  • Lab 1 (36 m2): Hemi-Anechoic Lab. Measurements, room acoustics, diffusion, instrument directivity, transducers spatialization, and room perception in listeners.
  • Lab 2 (36 m2): Critical Listening Lab. ITU standard room with multichannel and stereo audio, critical listening, evaluation, and technical ear training research.
  • Lab 3 (28 m2): Performance and Recording Lab. Recording of individual performers, motion capture measurements of respirology and performer movements, virtual reality applications.
  • Lab 4 (23 m2): Audiovisual Editing Lab. Preparation of test materials, PR and demo preparation, content for archiving, viewing/listening space.
  • Lab 5 (52 m2): Multimodal Shared Reality Lab. Audio, video, vibration capture and display and presentation, broadband, high-resolution network communication space, ultra-videoconferencing.
  • Lab 6 (23 m2): Perceptual Testing Lab. Experimentation on music and sound perception (will also double as an additional audio editing suite).
They have some very nice resources and facilities.

From the CIRMMT webpage: The CIRMMT community is interested in interdisciplinary research related to the creation of music in the composer's or performer's mind, the performance of music, its recording and/or transmission, and the reception of music by the listener. It is also interested in the ways in which vision, haptics and touch interact with music and sound.  CIRMMT seeks to develop innovative approaches to the scientific study of music media and technology, to promote the application of newer technologies in science and the creative arts, and to provide an advanced research training environment. 

Wednesday Aug 01, 2007

I like to listen to music when I'm working. Often I will listen to music from Pandora or using their flash-based players.  This is great when I am processing email or browsing the web, but if I am doing serious work, especially when I am in Netbeans, none of the flash-players seem to cut it.  It seems that whenever Netbeans is thinking hard (which is almost all the time with its new context-sensitive editor)  every flash player seems to stutter and skip.  The listening experience goes down the drain.  I suppose if I had one of those new-fangled dual core laptops I'd be in good shape, but with my 2 year old Ferrari, I have to use a more traditional desktop music player like xmms to get a good, consistent listening experience from my laptop. 

Sounds like it is time for an upgrade, now to convince my boss.

 Update:  What laptop do I have?  It's and Acer Ferrari 4005, running Ubuntu Edgy Eft.


Getting a cascading style sheet to work properly to work properly is always difficult for me.  Sometimes (most of the time) things just don't work they way I think they should.  This problem gets magnified when working on a web app - the edit/build/deploy/test cycle can take 60 seconds or more, add to that the problem of the browser or web server occasionally caching the CSS so I don't even see the changes that I thought I made and it can get downright frustrating.

 Coming to the rescue is the Web Developer Toolbar.  This firefox plugin adds a toolbar with lots of nifty tools useful for developing and debugging web apps.  The toolbar has some really useful features for debugging CSS.  Turn on "View Style Information" and whenever you mouse-over an element in your browser, it put a box around it and shows  you the style information like so:

html > body > div #banner > div .bannerStatusBox > table .bannerStatusBox > tbody > tr > td .bannerLeft

This can be very helpful if you are using a toolkit like GWT that attaches its own stylenames to widgets.

Perhaps the most powerful tool in the Web Developer Toolbar is the CSS Edit feature.  This feature allows you to live-edit the CSS of the current page. Make a change to the CSS and instantly see how it affects the page.  The edit/build/deploy/test loop drops from 60 seconds to 0 seconds.  It is an extremely useful way to debug and experiment.  When you are happy with the updated CSS, you can save it back to your project directory, so that the next time you do the formal build/deploy your CSS changes will be there. 

 There are lots of other nifty features in the Web Developer Toolbar. If you are doing any web development, I don't see how you can live without it.

Monday Jul 30, 2007

In this forum thread, Robert Kaye (in charge of mayhem at MusicBrainz), says that a new release of the MusicBrainz music metadatabase will include support for social tagging of music.  That will a great addition to dataset, but I imagine that it will take along time before MusicBrainz has built up as rich a set of tagging data that has accumulated.   Hopefully, the MusicBrainz community will  make it easy to tag by incorporating MB tagging seamlessly into various music players.  It'd be interesting to see how these tags differ to the types of tags that are being applied by and qloud users.

Friday Jul 27, 2007

In Sun Labs, there's a gaming project and a music project ... so naturally we have a lab where we can study the ultimate synergy of gaming and music: guitar hero.

When I'm working from home, I don't get to have my GH fix testing time, but I can at least satisfy the need to listen to some guitar hero tunes using tag radio:


This blog copyright 2010 by plamere