Sunday, October 09, 2005

Running fluff and linguistics fluff.

This morning I did this race:

This was the sixth time the race took place and the fourth time I've run it; it turns out that I've been in Seattle a while now, long enough to have stuff that I do every year, stuff that hasn't been around the city that much longer than I have. In any case, it was a fun race as usual. I expect my chiptime will be about the same as my time at the RTC race last summer, about 23:30, or about 7:30 per mile. (Yes, a race that does chiptiming at the finish and at the start, an increasing rarity around these parts.) I'm happy with how I did, but when I get back from vacation I'm going to shake up my training regime some. Once and for all I need to bite the speedwork bullet and get it done. I think I can be significantly faster if I train right.


Other things: Some of you know that Chris and Carlton have been studyng Japanese for almost a year now. It's led to some interesting discussions around our house about syntax, pragmatics, and language acquisition, which topics I haven't worked on in direct contiguity since finishing my dissertation, and most of all about language typology, which I've been thinking about quite a bit lately anyhow because of some things I'm working on.

It's one thing to identify the principles according to which languages vary and to demonstrate that all human languages are in their most essential properties the same, and formal linguistics as a field has been pretty good at that in the last decade or two. It's quite another thing to take some principle according to which language varies and state it elegantly and precisely enough that someone can implement it in software in a way that will produce appropriate, expected behavior for the range of languages that the software must support. It's easy as a formal linguist to be satisfied with a theoretically plausible but ultimately not very practically extensible theory. Computational linguists don't really have that luxury.

That isn't to say that theoretical constructs aren't useful, because they can be. But I can feel the changes in my thinking as I continue to move towards a more computational and somewhat less purely formalist style of thought, largely because of the work I'm going to be spending most of my days doing in the coming several months. A few years ago I was all formalist, not only in the topics I considered but also in the topics I considered worth considering. Now I don't really have that luxury, and I'm glad.

It was interesting hanging out with Elsi in LA and weighing her reactions to various topics that I'm thinking about, even though we didn't have a chance to discuss much linguistics. It served as kind of a placemarker for me, a Platonic form of something (because you don't really get any better at this particular form than Elsi is) that a few years ago I really aspired to be. For that reason it's also a marker of the ways and directions I've shifted -- neither better nor worse, but it's definitely different. And not especially in the ways a formalist, particularly an academic formalist, would expect, but rather in other respects I wouldn't have begun to suspect a few years ago.


Post a Comment

<< Home