Duke Listens!: Visit my main blog at MusicMachinery.com

This is my last post to the Duke Listens! blog. I've been posting to this blog for nearly 5 year starting with my first post on May 3, 2004. At that time, I was still working on speech engines, and the blog name was "Duke Speaks!". A month later, I wrote my first post that hinted at my interest in music information retrieval.

Blogging has been good to me. It serves as my backup memory. My nearly 50-year-old neurons may not remember everything, but I can usually remember if I've blogged about something, and a re-reading what I wrote refreshes my memory. So, for instance, if I can't remember the difference between the Idiomag API and the YES API, I can just read what I wrote.

Blogging has been a great way to get connected with folks. When I write something about a new Music 2.0 startup, I'll often get an introductory email from the CEO or CTO of the company the next day - I have a pretty good virtual rolodex due to my blogging.

Blogging has helped me improve my writing. Writing something every day that will be read by a thousand or so people has been a good exercise. I'm not sure if my writing is any better than before I started blogging, but it certainly is faster. I used to have a real problem with writer's block (and I still do when it comes to writing papers), but now blog posts seem to flow much more easily. Blogging has also helped me improve my thinking. By writing down my ideas on a topic I'm forced to think things through.

Blogging has given me a thicker skin. On the Internet, people like to disagree and criticize -especially anonymously. I've received my share of criticism, some gentle, some harsh and some downright offensive, but after 5 years, the critics don't bother me to much, and I can just reject and delete the offensive ones.

Some stats for the blog:

Total posts	1079
Total comments	1143
Daily uniques	~2,000
Google reader subscribers:	> 300
Where readers live	55 different countries

I want to thank all of my loyal readers for visiting Duke Listens!. I especially want to thank the many folks who've contributed ideas and links. I especially want to thank Steve, Jeremy, Zac and Oscar who have been particularly faithful in providing me with blog fodder.

And so with that, I write my last post of Duke Listens! I hope you all will continue to follow me on Music Machinery. Duke Out!

Tuesday is my last day at Sun. After 9 years, I've decided to move on to the next thing. During these last 9 year I've worked on some really fun projects and worked with some really smart, creative and interesting people. The work environment at Sun Labs is really special and I doubt that I shall ever experience anything like it again. Bob Sproull, the head of the labs believes that the best way to run the labs is to let the researchers work on the things that they are passionate about. Some companies have a practice called 'Friday Projects' where you can spend one day a week working on a project that you find interesting. For us in Sun Labs, every day is Friday.

With this freedom, I've spent the last four years working on the problem of music recommendation and discovery. I've written lots of code, given lots of talks, presented demos, written papers, filed patents, gave interviews for reporters, dined with an astronaut, written lots of blog posts, and listened to lots and lots of music.

I've also made many friends, and saying goodbye is the hardest part of leaving, but I suspect that many of our paths will cross again.

The labs is an extremely comfortable place to be. I could easily stay here for the rest of my working career and be quite happy and productive, but I have an itch to go and try to do something really great. I think I've found the place to go and do it, and so Tuesday is my last day at Sun, and Wednesday starts my new adventure. Feel free to follow along on my adventure.

Check out my review of the new music app for Twitter - Twisten.fm over at MusicMachinery.com

Oscar's Thesis - ready to read, check out the details on MusicMachinery.com

I've pushed out a new revision of the SXSW Artist Catalog. Details are posted over at Music Machinery.

The Beat Crave published recently published their list of the top 5 most overused animals names of indie rock. I wanted to see if this was generally true, so I grabbed a copy of of WordNet, a list of the top 20,000 most popular artists, and wrote a little code to find out the answer. I posted my results over at the Music Machinery.

Blip.fm is often described as a twitter for music. Blip users post ‘blips’ to tracks - and as with Twitter, others can follow your Blips and listen to what you’ve posted. It’s music-micro-blogging. The folks at Blip.fm have built some web services around Blip.fm that let a 3rd party application do all sorts of nifty things with Blip. I've taken a look at the Blip.fm API and have posted a review on the Music Machinery blog. So go check it out.

The Sphinx-4 team has released 1.0 beta2 of Sphinx-4, which is a state-of-the-art, speaker-independent, continuous speech recognition system written entirely in the Java programming language.

New features and improvements include:

• new much cleaner and more robust configuration system
• refactored to make use of java5 language features
• new API to setup speech recognizers without any xml
• improved feature extraction (better voice activity detection, many bug-fixes)
• Cleaned up some of the core APIs
• 'include' and 'extends' for configuration files
• better JavaSound support
• fully qualified grammar names in JSGF (Roger Toenz) support for dictionary addenda in the FastDictionary (Gregg Liming)
• added batch tools for measuring performance on NIST corpus with CTL files
• many performance and stability improvements

Taylor McKnight 'spent an hour playing around' with the CSS for the SXSW Artist Catalog. He's cleaned up the design and added a dash of color. I really like the new look. Thanks Taylor!

One way to make sure that your band gets the most space on the marquee at SXSW is to have a really long band name. Some bands have done just that. Here's a histogram of the name length of the 1200 or so artists that are appearing at SXSW:

This histogram shows that the most frequent band name length is 11 characters, (the average is 12.9 characters) with some as short as 3 characters, but there are quite a few with lengths of 30 characters or more, with one topping out at 43 characters.

Here are the bands that are getting the most space on the marquee:

• Jon Langford And The Pine Valley Cosmonauts
• Johnny Goudie and The Little Champions
• Samantha Crain & the Midnight Shivers
• Someone Still Loves You Boris Yeltsin
• Solange and the Hadley Street Dreams
• Edward Sharpe and the Magnetic Zeros
• Hermano L backed by Ranking Soldiers
• Danny And The Champions Of The World
• I Love You But I've Chosen Darkness
• Bryan Scary and The Shredding Tears
• Steve Goldberg and the Arch Enemies

And here the bands that are getting the least space:

• WAZ
• XYX
• TAT
• Mom
• Fun
• fiN
• CLP
• Cla
• CKY
• Blu
• Aun

Note to bands - when you are picking the name for your band, try googling for it first - it is going to be a lot easier for people to find you with a name like "Jon Langford And The Pine Valley Cosmonauts" than for a band with the name "fun"

There are so many artists going to SXSW, it is hard for me to decide which bands that I'd like to see. I know I'd like to catch a band that's on the verge of breaking out - a band that just might make it big in 2009. To help me find the potential break out bands, I've updated my SXSW Artist Catalog to also list the 'Rising' artists. . These are the artists that are getting a lot of buzz in the mp3 blogs, but don't have a whole lot of listeners yet. Presumably these are the bands that are getting noticed by the critics, but haven't built a listener base yet.

According to this metric, the band to see at SXSW this year is Come On Gang! - they have a 'hotness' score of 28, while at Last.fm they only have 45 listeners.

Much thanks to Chris Gaines for suggesting this metric.

The presentations for RECKED, are now available online. They are all pretty interesting.

I was using the nifty developer's API at the Echo Nest to fill out some of the data for the SXSW Artist Catalog. I encountered a bug where the API would return an HTTP response code of 500 for some artists. I posted a description of the bug, grabbed another cup of coffee and then re-ran my artist crawler. I noticed that my crawler was no longer encountering the bug. Suspecting that I may have just encountered a glitch, I went back to the Echo Nest developer's forum to add a note when I saw that they had already replied to my request and had fixed the bug. So in four minutes they had: 1) Noticed my bug report, 2) identified it as real bug, 3) fixed the bug, 4) deployed the fix, and 5) Posted a response to the bug report. I think that may be a world record bug fix time for a deployed web application. Nicely done, Echo Nest.

I've updated my SXSW Artist Catalog to include a list of the hottest bands attending SXSW. These are the bands that are getting the most buzz in the music blogs. I'm getting this hotness value from the musical brain at the echo nest via their developer APIs.

Monday I shall be in Montreal, giving a talk about music recommendation and discovery at Doug Eck's Music and Machine Learning course (or more properly, 'Musique et Apprentissage Automatique'). Also joining us will be the students taking the McGill course "The Remonetization of Music and Other Promiscuously Transportable Media Objects".

To get people thinking a little bit about the issues surrounding music recommendation and discovery, I offer these two pre-class assignments:

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

• Last.fm
  
• Ishkur's Guide to Electronic Music
  
• Hype Machine
  

The second assignment is to actually make a music recommendation. Here's the details: I'll be attending the SXSW Music festival in March where about 1,200 bands will be playing. For your assignment, make a few recommendations for me of bands that that I should see. Post your recommendations and how you came up with them as a comment to this blog post. The SXSW Band list is here: Official SXSW Showcasing Bands. You can perhaps glean something about my music taste by looking at my Last.fm profile.