MIT EECS | Hudson River Trading Undergraduate Research and Innovation Scholar
LEXI Project Automatic Speech Recognition
- Natural Language and Speech Processing
Most current approaches to speech analysis focus on machine learning with large amounts of training data, and do not model human speech using linguistic theory. The aim of the LEXI project, which stands for Linguistic Event Extraction and Interpretation, is to build a model for speech analysis using new linguistics research by separating acoustic cues and bridge the gap between linguistics and machine learning. The goal is to build a system that can take a speech signal, find the acoustic cues, use the cues to find distinctive features and phonemes, and finally convert these phoneme patterns into words.
“I major in computer science and minor in linguistics, so speech analysis is very interesting to me as it is a natural combination of my two favorite subjects. I’m excited to use my skills as both a computer scientist and linguist in this SuperUROP to participate in interesting research and further improve the project.”