I always wondered why University of Melbourne didn’t pony up the cash to keep Dipesh Chakrabarty or even why my alma mater couldn’t hold on to to it’s most brilliant profs (but that’s a different story. properly titled, “Why UR chose aesthetically pleasing architecture and painfully perfect shrubberies over scholarly substance and exciting research.)

very enjoyable anyhow.

Based on a v quick read through his paper on the project (link), it seems Deb Roy has his own model for language acquisition in humans. The basic idea seems to be that children learn by mapping objects and routine actions to different parts of speech. He has built mathematical models which have been able to learn nouns and verbs from videos of infants playing with their mothers. He now wants more complete data to conclusively prove his hypotheses.

Chomsky’s thesis has been widely discredited in the field of linguistics It has? As a lay reader, thought it had gained even more credence recently…

One of the byproducts of the project might be evidence to disprove Chomsky’s hypothesis(though it cannot prove it). If he can get a program to learn a language based simply on statistical models, without using an inbuilt grammatical model, it is a setback to the idea that language needs to be inbuilt. From the paper:

A critical question underlying any model of learning concerns the balance between nature and nurture. HSP brings a new perspective to this age-old debate. Given a near-complete, contextually-rich record of a child’s first three years, what are the set of ontological constraints that must be built into a model for it to successfully learn aspects of language? If a machine can be shown to acquire some capability or structure X without corresponding innate preconditions, this provides evidence that the child’s environment provides X – and thus need not be innate.

let me soak it in for a while.

mmmm. good.

I guess that people who work don’t read stuff like this.

SudoSecularSAsian: All is forgiven. This is the brave new world of lexico-typographical expressivity! Au revoir!

my favorite part. mostly because and probably sum up the sum total of my internetese and i’m often afraid and embarassed to ask what all those acronyms people litter their posts with mean.

Or has Sepia doing some market research – and found that majority of site visitors are ivory-tower dwelling effete, pointy-headed reification-fetishists – hence the degradation to suit the “target audience”?