Research

Thoughts

I'm interested in...

• understanding language
  what drives the evolution of form?
  how can we capture representation of meaning?
• modeling language using computers
  what are the current capabilities of language modeling tools?
  how can we evaluate them in comparison with human language perception?
  what can computational models tell us about human language evolution?
  
• designing language technology
  how can we improve current language processing methods to work for low-resource settings and non-conventional communication?
  at the social level, how can we actually employ language technology to reduce communication barriers?

Current

Studying individual neuron responses to speech

Current questions: Can artificial neural network encodings model brain responses to speech? What aspects of the ANNs contribute to modeling ability? How is the neural encoding of speech information distributed?

CCN Conference Abstract

Future questions: What is the role of probabilistic computations in speech processing, and how do they manifest at the neural level? How does information flow through the brain networks that process speech and language? How do neural circuits distinguish speech from background sound, comprehensible speech from incomprehensible speech, and so on?

Past

Translating English instructions into temporal logic

Robotic planning algorithms let robots find optimal & safe ways to solve formally defined tasks. However, formally defining a task requires domain-specific knowledge and may be inaccessible to the people actually using the robot. In this project, we tried using NLP models to fill in the translation gap between a colloquially described task and a formal logic description.

MEng Thesis // EMNLP Conference Paper // MIT News coverage