In the LEXI (Linguistic Event Extraction and Interpretation) speech analysis framework, acoustic cues are patterns in the speech signal that can be used to infer abstract categories such as distinctive features of phonemes. This relationship can be a simple one-to-one mapping, but more generally, it can be a many-to-many mapping. This project will build the framework for the interpretation process from acoustic cues to distinctive features, focusing on using VOT (Voice Onset Time) to infer the LEXI distinctive feature [spread], which will map to the linguistic distinctive feature [voiced]. The project will use acoustic cue-labeled databases of English in the Speech Communication Group, and in the future, generalize to be compatible with other languages.

My SuperUROP project allows me to learn about the phonetics-phonology interface through the construction of a bottom-up framework for speech analysis. The long-term exploration that an opportunity like SuperUROP provides will help me familiarize myself with each step of the interpretation from speech signals to language. This includes the analytic methods that I am excited to learn, and also my theoretical knowledge from linguistics classes that I hope to apply and expand upon.