The most interesting part of Andrew Carnie’s “Introduction to Syntax” for me was his discussion of Universal Grammar and different theories about language acquisition. It seems natural that, amidst Carnie’s discussion of the immense efforts linguists have made to scientifically pin down the syntactical rules of language, how all of us somehow naturally manage to accomplish this effortlessly in the relatively short amount of time it takes us to learn how to speak. As Carnie establishes midway through the first chapter, language has a recursive structure that allows it to create infinite permutations, which makes the task of extrapolating syntax rules from language around us a daunting one in terms of size. In response to the idea that the language data set we as children are exposed to is too small to allow us to adequately grasp these rules, Noam Chomsky’s theory of Universal Grammar posits that there is an innate human facility for language--that to some extend, these rules are innate to us.

What I found most interesting about this idea was its implications for machine learning and natural language processing. In Symsys1, we talked a bit about the recursive structure of language and how ‘context free grammars’ create recursive, hierarchical syntax trees to generate input. Given the amount of overlap between the two classes, I started thinking a bit about the problems each raised for the other. If Universal Grammar is right in positing that our ability to learn language is dependent on some innate facility, would it be possible to ‘teach’ a computer that facility, or is it something dependent on our biology (and can those structures be reproduced within a machine)? If neither is the case, what does that imply about the nature of language? From the other end, thinking about all the efforts computer scientists have hitherto made to attempt to get machines to simulate real speech made me think about what insights computer science could bring to linguistics. Could attempts to teach computers language shed light on different models for language acquisition in children?