Alisa Liu

Northwestern University

Hi! I am an undergraduate at Northwestern University studying computer science and math, and I identify with the natural language processing and computer audition communities. My current research work includes definition modeling for word compounds, generating natural language captions for audio, and some tinkering in music generation. I’ve recently been thinking a lot about developing automatic measures of model confidence and how they can be used to enable different training techniques and incorporate human input. I have been super fortunate to work with Professor Doug Downey, Professor Bryan Pardo, and Dr. Prem Seetharaman, who have been the kindest of mentors and completely formative to my intellectual interests.


  • Natural language processing
  • Machine learning
  • Machine processing of music and audio


  • BA in Computer Science, Mathematics, 2020

    Northwestern University

Research Projects

Audio captioning

Generating natural language captions for audio clips

Augmentative generation for Bach chorales

Using heuristic evaluations of generated music to train a Bach chorale generator on high-quality generations

Definition modeling for noun compounds

Developing a neural language model that generates definitions and paraphrases of noun compounds (e.g. “caramel popcorn”)

Ensemble model for audio source separation

Built an ensemble model for audio source separation that can handle mixtures whose source domain is unknown, using a confidence measure …

Common sense QA dataset

Introduced an adversarially generated commonsense question-answer dataset

Analysis of error types in multi-sense definition generation

Evaluated the settings under which a multi-sense definition modeling system succeeded and failed


Model Selection for Deep Audio Source Separation via Clustering Analysis

Ensemble model for audio source separation, using a confidence measure to mediate among domain-specific models

Multi-sense Definition Modeling using Word Sense Decompositions

Definition generation for multiple senses of a word

CODAH: An Adversarially Authored Question-Answer Dataset for Common Sense

An adversarially-constructed dataset for common sense QA, collected from Northwestern ML students!

Comparison of Discourse Surrounding CRISPR/Cas9 in the Media and Peer-Reviewed Literature

Comparison of the portrayal of CRISPR/Cas9 in mainstream media and academic literature. (From my pre-med, baby-research days.)


CS 336: Design & Analysis of Algorithms

peer mentor, with Professor Jason Hartline (Fall 2019)

EECS 349: Machine Learning

peer mentor, with Professor Bryan Pardo (Spring 2019)

EECS 396/496: Statistical Machine Learning

course developer, with Professor Han Liu (Winter 2019)

CPSC 121: Models of Computation

undergraduate TA, with Professors Alice Gao, Steve Wolfman, Ryan Vogt (Term 2 2018)


Some study guides I've made

I really enjoy making study guides! Here are a couple I'm especially proud of, and I hope they are useful resources for other students. …