Research explained
Published: March 21, 2008Tags: phd
In my last entry I said I'd explain the research page page which mysteriously appeared during my site redesign. Here's the story.
My first job at m.Net Corporation was basically to refine and extend some work done as part of a joint research project between m.Net and a research psychologist from my alma mater. This psychologist was Daniel Navarro, an insanely smart guy who, despite being psychologist, actually understands things like maths and statistics and can even write code (though be fair his code sometimes sucks).
Working together we had moderate success in adapting latent Dirichlet allocation, a mathematical model originally developed for natural language processing, to a collaborative filtering problem as part of m.Net's customer analytics research. It was pretty cool stuff, and I learned a lot. I was genuinely surprised and excited to realise that some psychologists actually do things like heavy Bayesian statistics and intense number crunching, instead of just blindly assuming that all the world's data is normally distributed and interpreting simple linear regression as the Word of God (which is what mathematicians generally assume psychologists spend all of their time doing - it's a reputation they deserve for teaching their students from a book called Statistics Without Maths for Psychology. I mean, really). Check out MIT's Computational Cognitive Science Group for a better idea of the cool kind of stuff some people do. Anyway, about a month ago Dan mentioned to me in passing that an internal PhD scholarship in the School of Psychology may be about to become available, and suggested that if I were interested he could try to convince them to let me apply for it, on the grounds that teaching a mathematician the basics of psychology is about 100 times easier than teaching a psychologist the basics of mathematics and hence recruiting mathematicians is actually a smarter way to produce good research in mathematical psychology. I said I was interested, because I really did find the LDA work I did interesting, and he said he'd try but that I shouldn't get my hopes up because it was a long shot. So I didn't. Fast forward to earlier this week and the last bit of paperwork has gone through and the scholarship is mine. Sometime before the end of the month I'll be starting a PhD, with Dan as my supervisor. The topic of study has not yet been finally decided, but it will revolve in someway around how humans firstly learn and subsequently understand language (and these are obviously related problems, considering the way that advanced language is typically learned via explanation using simpler language) and involve as much mathematical modelling and number crunching as I can possibly squeeze into it. I'm very excited about possible applications of these models, to things like improving the "intelligence" of tools like search engines and news aggregators and, perhaps more ambitiously, using software to "bootstrap" the semantic web by auto-generating RDF files en masse.
So you can expect any papers or the like that I write in the course of this PhD to appear on my research page, any software I write as part of it to appear on my software page (under a BSD license, of course), and occasional thoughts to appear in this blog.
This doesn't explain the fact that all of this is happening with probability 0.5. I'll leave that for another entry.
Oh, and I am trying to arrange to stay on at m.Net for one week a day during the PhD, because it's an awesome place to work and I'd be genuinely sad to leave for good.