In recent years, there has been an increase in demand for, as well as innovation in, speech-analysis services. Unfortunately, many of the techniques and services are dependent on high volumes of quality data to perform well. My research with the MIT Speech Communication Group aims to solve data dependency problem for development of speech analysis system. We combine linguistic knowledge and computer science techniques to train and evaluate modules that recognize linguistic contrastive parameters in languages. From the detection results, we can use linguistic constraints to infer intended distinctive features and subsequently recognize phonemes and words from the extracted speech signal.

I joined this project in my sophomore year in college. I really like the linguistic aspect of the research and am particularly interested in working with machine learning. I hope to continue learning a lot of technical skills and research skills through SuperUROP. I am very excited to see the result at the end of my research project.