December 12th, 2005

beartato phd

(no subject)

In this language log post Mark Lieberman links to a longer description of an interesting experiment in modeling vocabulary creation. His conclusion is that for a language community to successfully negotiate consensus about how words are pronounced, it is necessary and sufficient (given some other assumptions about how pronunciations are learned) for the individuals in that community to have a continuous ("gradient", "probabalistic") internal representation of belief about pronunciation, and a discrete ("categorical") way of perceiving pronunciations.
beartato phd

(no subject)

Major twelf success just now --- the structure of this proof was like 5 stages to begin with, and ever since starting to formalize it each stage has a nasty habit of needing to be dividing up into zillions of little substages. Moreover, it is hard to tell in advance which substage will be the really hard one that I need to think the most about. So I just blunder forward a lot of the time, fixing problems whenever they come up. This means the structure of my approach to, say, stage X looks like this: Collapse )

Anyway I've said it before, but it sure competes well with the excitement of watching like action movies or sports or whatever when I'm all furiously typin' away, and I'm all getting towards the conclusion that I've been working for for a few days, and I'm like "Server OK? Server OK? Server OK? SERVER OK!" The huge, huge mess of coverage errors that I had to wade through to get to that point only made the partial victory sweeter. I'm particularly happy that I found a solution to a nasty termination problem, via tossing in another argument to one typing definition that tracks its "shape".

---

But now I think I've set a new personal record for most complicated set of coverage errors elicited by a single %total declaration: 180 missing cases, for a total of 2,638 lines of error messages.
beartato phd

(no subject)

Too many things to do and not enough time to do them in!

(1) Had dinner with wjl and platypuslord at Veracruz. (2) Since grammarnerd recently mentioned seeing Brad Yoder play, it reminded me that I hadn't been to one of his shows in a while, so I made at the time I read that post a mental note to see tonight's show, which I did, and wjl came along as well, having never heard of Mr. Yoder before. It was a good show; I really love the depth that Jason Rafalak's bass playing adds to the songs, and of course they are good songs to begin with. Heard a bunch of familiar stuff like Ginko Leaves and Love is All I Have for You and Skyler and Head Together. I was a jerk and left early, but I had an excuse, which was (3) to go meet up with subpar and lambdacalculus at the appointed time for the Last Monday Night Nachos. "Last" since neal is all graduatin' and movin' back to new mexico soon.