I am a Ph.D student at the Language Technologies Institute, a department of the School of Computer Science at Carnegie Mellon University. My current work is to develop new technologies in order to build more effective automatic speech recognition software. This is an important area of research because spoken language is perhaps the most natural mode of communication for people, but so difficult for computers. Among other reasons, automatic speech recognition is hard because computers, telephones, and devices have no faculty for understanding meaning in spoken language.
My work is on figuring out how to encode some of the meaningful information that may help guide automatic speech recognition. For example, “The pulp will be used to produce newsprint” and “The Pope will be used to produce newsprint,” are acoustically very similar. But to most people the former seems much more plausible, simply with a little common sense. My work is on how to encode and use some of that common sense in an ASR system.
I collaborate primarily with my research adviser Scott Fahlman, who specializes in how to represent knowledge in computer software, and Bhiksha Raj, who specializes in automatic recognition of spoken language.
In addition to speech recognition, I have research experience in natural language processing, knowledge representation, machine learning, information extraction, and information retrieval. Before coming to CMU in 2005, I completed an MS at University of Illinois Urbana-Champaign (thesis adviser: Dan Roth), and my BS at University of Massachusetts Amherst (thesis adviser: Andrew McCallum). In my spare time I enjoy softball, Mozart, and theater, among other things.
I proposed by Ph.D. thesis in December 2009. Here is the proposal: “A Knowledge-Base Architecture for using Semantics in Automatic Speech Recognition.”