The overall goal of this project is to aid in the pursuit of improving speech recognition models. I will design a data structure of human speech, combining prosodic and acoustic cues. This data structure takes as input an annotated speech sample and outputs an organized hierarchy of information. Significantly, it creates the context needed to compare the predicted cues of a speech sample with its realized cues. Thus, my data representation can be used as the input for a tabulation program, allowing for the number of predictions vs. realizations to be calculated. These results are then used as the input for a post-processing program which compiles all possible contexts that can arise in a speech utterance. Combining these components together creates the Cue Production Profile system.

I am participating in this SuperUROP because I am eager to pursue research in an area that combines my two major interests- language and technology. I have been with the Speech Communication group for two years, so I am excited to continue working with them under this new research position. By the end of the program, I hope to publish a paper and present my results at a conference.