Register Recognition
As a Linguistics and Computer Science major, I tend to think of language questions in terms of computers and find computer questions that have to do with language. For Computer Perception, I chose to design a program to distinguish between formal and informal registers of speech. That is, our speech changes depending on our audience and I want to see if we can detect that automatically.
Specifically, I'm trying to classify samples of African-American Vernacular English (AAVE), some of which are more formal, almost Standard English (SE). I first got the idea to try this when I wanted to analyze some linguistic data but couldn't tell which data was pure AAVE and which had influences of SE without picking apart every word and inflection in the sentence.
Now I've written a system which comes close to distinguishing the two, without relying on the knowledge of a linguist. It turns out intonation is the big give-away. My system estimates the change in pitch across a sentence, statistically analyzes that change, and uses it to distinguish test data, effectively telling us which rhythms are natural in AAVE and which aren't.
|
|
Anne-Marie Frassica '09
|
|
