I am a PhD Candidate in Linguistics and a data science instructor at UC Berkeley.
My research focuses on understanding the methods and practices in acoustic phonetics.
I am particularly interested in quantitative comparisons of the quality of different
representations, and understanding
what factors influence the emergence of best practices in phonetic methodologies.
In my research, I also explore alternatives to traditional phonetic features in
acoustic analysis, and draw on other fields of acoustics and other adjacent fields
to identify features of interest.
I have formal training in statistics, data science and machine learning, and have
been teaching data science to social and other scientists since 2019 as a part of the UC Berkeley D-Lab.
I believe that
quantitative foundations empower scientists to address interesting and important questions
in their areas of expertise, and am passionate about expanding access to these skills.
A key part of any phonetic study is how acoustic phenomena are represented.
This project focuses on quantitative comparisons of the quality of representations
and understanding the social and commmunity aspects that influence adoption and choice of acoutic parameters.
I also explore novel parameterizations, often drawing on broad acoustic fields.
Data science and pedagogy
Data science methods have become more and more central to social science research, but training in these
methods is not yet common practice. I am passionate about helping scientists expand their access to
data science and computational methods through formal and informal instruction.
Unsupervised vowel clustering
Machine learning methods show promise for supplementing traditional analyses.
This project explores alternatives to the ubiquitous K-means for vowel clustering.
Collaboration with Jennifer Kuo (UCLA)
Probing neural net representations
Neural nets have a significant ability to detect patterns in data,
and may be informative of the kinds of patterns that are distinctive of phonetic contrasts.
This project probes representations at different levels of neural networks.
Survey Builder in Python
Code for a light-weight command-line program to build and run methodological surveys.
Introduction to bag-of-words and sentiment analysis
Gentle introduction to text analysis and representations for non-major computer science students.
Vowel Clustering Methods
Project exploring OPTICS as an alternative to the common K-Means clustering alorithm for vowel space description. Collaboration with Jennifer Kuo (UCLA)