Undergraduate Honors Thesis in Linguistics
(Published on: May 22, 2009)
This is the last thing that I wrote as an undergrad at Cornell, and is probably the most significant academic project I took on in my four years there. I started it later than prudent, because I wavered for a while on whether I really wanted to do it, but in the end, I'm really glad that I did. For that, I thank my major advisor Abby Cohn for pushing me to do it in the first place and my thesis advisor John Hale for being a really smart guy with a personality to match.
Here's the abstract:
Self-paced reading studies are a valuable tool for gaining insight into the domain of sentence comprehension difficulty. Since these tests directly measure the difficulty that a human reader experiences during the comprehension of a sentence, it follows that any model which claims to be realistic must also produce reading times that are consistent with empirical data. To this end, I propose a mathematical framework to describe the mapping from parsing models to reading times and provide a software implementation of this framework. I then use this implementation in an experiment to guide the search for the correct framework functions based on the known performance of a Left-Corner parsing strategy. Next, I apply the framework individually on Top-Down and Bottom-Up parsing strategies as negative evidence to further support its validity; a brief digression is then taken to explore and qualify the higher-dimensional properties of the framework. Finally, I show that the proposed framework is general enough to capture pre-existing theories in the field.
