project details

To achieve my goal of using NLP to better understand how each person talks, thinks and feels, I have worked on

During my MS Computational Linguistic (NLP) at University of Washington

  1. How personal attributes (such as one’s favourite food and music) can be extracted and inferred from conversational language, and their applications in open-domain chit-chat agents and multi-task task-oriented dialogue. PDF Code

  2. How to use commonsense knowledge and language models to understand how humans form expectations of events unfolding in stories and when such expectations fail? PDF Code

During my SDE Internship at Amazon Alexa AI Knowledge

  1. What can we do to link natural language expressions to objects and concepts (i.e. entities) in the real world? How can we support machines to make use of this world knowledge in understand natural language, as humans do? Publically Released Product

During my BA Psychology and Education at University of Cambridge

  1. What characterizes empathetic language, as used in social media (Reddit)? How can empathy be efficiently and accurately labelled based on signals from social interaction? PDF Code

  2. Why does one movie character remind you of another totally unrelated movie character? What can such thematic similarities (also known as tropes) tell us about similarities in human experiences? PDF Code

  3. Can personality traits of students be predicted based on what they write in diary-like writings? Can we make good prediction based on writing from 1, 2 or 3 years ago?

  4. What do people associate with their subjective well-being (or happiness)? Can we find that out based on statistical trends in their real-world language use? PDF

  5. How to help people find others who have had and written about similar personal experiences? PDF

  6. How can we help students to be more interested in science by tapping on their natural curiosities and interest outside of science? How can we represent and learn such interests using natural language processing? Website

  7. How can we offer personalized guidance to students 24/7 when they encounter a question that they’re stuck on? Might retrieving semantically-related questions with step-by-step solutions help? Website