I went to my first "proper" conference this weekend. (By proper I mean spanning multiple days, having more than a few in-crowd delegates, and based somewhere other than Norwich.)

It was Language and Music as Cognitive Systems at CRASSH in Cambridge. It was very different to the kind of academic discourse I've become used to in that it had the feel of a science conference. There were lots of tables of numbers, charts, and descriptions and justifications of experimental methods. I think I probably prefer the more verbose style of humanities discourse. Though, on the other hand, it is an interesting idea that work in the humanities could emulate some of the aspects of the sciences. For example, poster sessions are now common-place at humanities conferences but were originally a feature of the sciences and there is a conference coming up which makes this borrowing even more explicit: Empirical Musicology (I may even submit an abstract on computational something-or-other...)

This didn't detract from some of the content, though. The empirical style I guess is a consequence of the kind of issues which music and language share: mainly centred around sonic perception, the relationship between syntax and semantics, and propositional vs. expressive meaning. I was able to relate some of the cognitive issues to my research through the vast amount of work that has been done in modelling cognitive systems in musical perception and creation in computers (though these issues only form a "third" of my work).

Jamshed Bharucha's keynote on Saturday afternoon was very entertaining but I was particularly interested in Geraint Wiggins' response. He neatly highlighted all the concerns around computer modelling that I've been dealing with recently: comparing descriptive and explanatory models; not overstating the metaphoric qualities of computer models - they are far too precise and require a direct mapping to the semantics of the modelled object; arguing that models which are merely specific rule-bases (in the broadest sense) are not models, but notations; that supervised learning leads to treating descriptive models as ends in themselves; that unsupervised learning models can be used to generate hypotheses; and being able to view different levels of abstraction and to move an explanatory model from one level to become a descriptive model at another.

And, bizarrely, I met an old school friend! It was very strange. We can't have seen each other for six years and, suddenly, there she was. She's doing research at Cambridge.

HTML Course

So I taught the first session of my course on Practical Digital Publishing in Academia this morning. I thought it went quite well, probably because of rather than in spite of the very small turnout. (Apparently the current first year are notorious for poor attendance.)

Postgraduate Seminar Series

Did I tell you I'm running the postgradute music seminar series next year? Well, I am. I need to get the PGs to suggest speakers to invite and to submit papers themselves. Its a very difficult job. I'll have to set an example by giving a paper early on in the season and inviting lots of interesting people. The only thing is that I'll people who I think are interesting, but who may not be of much interest to anyone else. Ah well, if they don't ask they don't get, I suppose.

Right, I suppose I had better get on with my thesis.