The Semantic Web and Why You Should Care - a highly informal slidebook

As you may have noticed, my posting has been super light (even for me) for the last few months. Sorry about that. Anyway, the reader may be interested in the presentation I gave at the IEEE/ACM Joint Conference on Digital Libraries, Workshop On Integrating Digital Library Content with Computational Tools and Services this morning (about 4 hours ago actually). These slides really require the presentation audio to make sense and if it was recorded (unsure if it was...) I'll edit the slide share to have the audio track as well. Regardless, here's the slidebook. Have fun.

The Semantic Web and Why You Should Care.

View more Microsoft Word documents from Ben Fields.

back to playlisting

After a brief tangent into the wide world of mixing (of which I'll post some more in a bit from my proposed SMC paper in a bit) I'm back into playlist generation and related topics.  Along those lines it occurs to me that readers of my little blog may be interested in perusing my recently completed (Dec 2008 actually) M.Phil to PhD upgrade document.  For those of you unfamiliar with the British PhD system, PhD students start as a Master's of Philosophy by research student, then after about 24 months of independent research go through a process of summarizing and defending their work so far and what they intend to accomplish in the coming 18 - 24 months.  The outcome of this process is one of three things:

1. Your work is deemed interesting, rigorous and sufficiently likely to succeed in the next couple of years.  As a result, you upgrade to a PhD student and continue on with your research (this is what happened in my case)
2. You graduate at that point with a M.Phil.
3. You completely fail your upgrade process.

So that happened back in mid december.  Here's the abstract from my upgrade:

A framework is described to consider various real world playlist use cases.  Automatic playlist generation is introduced as a means to improve music recommendation.  Literature in related topics is discussed.

A sample of the Myspace artist network is examined to investigate the relationship between social connectivity and audio-based similarity.  Audio data from the Myspace artist pages is analyzed using well-established signal-based music information retrieval techniques.  In addition to showing that the Myspace artist network exhibits many of the properties common to social networks, it is seen that there is an ambiguous relationship between audio-based similarity and the social connectivity. Further the Myspace sample is examined with the pairwise relational connectivity measure Minimum cut/Maximum flow.  These values are then compared to a pairwise acoustic Earth Mover's Distance measure and the relationship is discussed.  A means of constructing playlists using the maximum flow value to exploit both the social and acoustic distances is realized.

Two playlist generation methods are proposed for development and experimentation.  The first is a direct extension of the myspace dataset analysis into a robust playlist system for interactive internet radio broadcast.  The second is content based system which uses expert constructed playlists to construct transition models which can then be used on new material.  This is followed by a discussion of evaluation needs and strategies. 

 If you're interested in reading the whole thing (comments welcome and encouraged) download the pdf.

On an unrelated note, I'll be a Yahoo Open Hackday 09 in Covent Garden in a couple weekends.  It's free and I believe there are still some tickets if anyone is interested.  It should be rad.

Conference alterations...

So this paper I'm trying to put together on mixing algorithms just didn't quite come together in time for DAFx. Or more exactly, the actual research wasn't quite done till like yesterday. But it's not all bad. I'm submitting to Sound and Music Computing 2009 instead, which looks like it'll be an interesting conference. This also gives me a couple more much need days to hash out a few more thoughts on things.
Here's a teaser from the paper:

For the purpose of this discussion we will be dealing with various types of song transitions in order of temporal complexity, from the simplest to describe in time to the most complex.

Song to song transition types


• Arbitrary Length Fixed Time Crossfade


• Phrase-Aligned Start No Tempo Adjust


• Phrase-Aligned Start, Running Beat Alignment


• Phrase-Aligned Start, Phrase-Aligned Finish

Once I've got this draft done I'll post some more bits. In the meantime, anyone have any further ideas on core subdivisions of song to song transitions, from the perspective of time alignment?

working toward a rigorous definition of 'continuous mixing'

So I have been thinking about this idea of continuous mixing, or generally speaking, more content aware automatic song to song transitions for some time now. To that end I'm aiming to put something together for DAfX on the topic. As this deadline for submission is fast approaching, my writing is generally focus in that direction. In the meantime however, I'll throw out a vague english definition of continuous mixing so that you the reader can get an idea of where I'm coming from on this topic, with some more rigorous definitions to follow.

continuous mixing: A song to song mixing technique where the goal is to obfuscate the transition between the two songs such that a casual listener cannot immediately pinpoint when the transition occurred. This can involve the use of beat matching/alignment, phrase matching/alignment, content aware equalization and other technical elements as well as sensible song selection with regard to harmony (i.e. key changes that make musical sense).

So that's my starting point. Anyone else have an opinion? What is continuous mixing mean to you? Bonus question! When is continuous mixing an appropriate transition technique in playlist presentation? I'm starting with modern electronic dance music, but these techniques can certainly apply to other musics (speed metal? maybe. free jazz? probably not...)

What is 'continuous mixing'?

So as some of you may have noticed iTunes 8.1 just came out. It contains something called iTunes DJ. There are two interesting claims it makes, one of which is good and the other, well, would be good if they did it. As near as i can tell (I've only played around with it for a couple minutes) iTunes DJ is basically Party Shuffle++.

The new thing that's rad is a new remote voting feature. If a bunch of people are at a party and they have iPod touch/iPhones they can vote for tracks to be included the ongoing playlist. This is a great idea generally (though I wonder what the avg house party ipod user % is like...). I've been playing with some other democratically driven social playlists, though mine are a bit less direct, involving voting on the destination of a small chunk of a playlist.

[rant]
The new that's not so rad is this claim about continuous mixing. The splash page states that "iTunes DJ automatically picks songs to make a continuous mix of your music." Since I'm working on automatic continuous mixing that got my attention. But the thing is they don't mean take a few track beat align them, overlap them, phrase align the crossfading etc. Apparently they mean constantly select some tracks using nearly random criteria (seriously, the second track it picked was a chapter from a book on tape that's in my iTunes library. It's even genre labeled ' Books & Spoken' so I don't know what sort of engine thinks that's a good track for the DJ playlist). I realize that at some level this a matter of language. Clearly whoever made the splash page doesn't have the same meaning in mind for 'continuous' that I do, which involves smooth mixing or at least removal of leading and trailing silence. But I guess I just think that if you're going to call a piece of software a DJ it ought to act like one.[/rant]

ISMIR tutorials

So I don't know exactly when this happened, but the ISMIR 2009 folks have published the tutorial schedule.  I of course already knew about mine and the music and the semantic web tutorial presented by Yves, Kurt et al.  The other 2 are 'MIR at the Scale of the Web' by Malcom Slaney and Micheal Casey, which I imagine will be all about using audiodb + LSH with extremely large collections of music.  I'm sure this will be a great tutorial so it's unfortunate that you won't be able to attend this tutorial and mine as they're at the same time.  In the other afternoon session you can attend 'Using Visualizations for Music Discovery' presented by Justin Donaldson and Paul Lamere.  As they both do great work I'm sure it will be an excellent tutorial as well.  Also, as a bonus, the tutorials happen to fall on the day of my birth, so there's that as well.  Kobe is shaping up to be a conference packed full of goodness...

beware the revolution

I joined twitter.  If anyone wishes to follow my thoughts an extremely terse form, this can be done here:

http://www.twitter.com/alsothings

Also, I added a feed in the lower right on this page.  As you may have already noticed.