Alice Lai

Senior Data Scientist, Machine Learning Department
Carnegie Mellon University

Pittsburgh, PA

Current work

I am a researcher in Machine Learning and Natural Language Processing working on applications of AI and ML to social good and public policy issues with Rayid Ghani at CMU.

Recent publications

Preventing Eviction-Caused Homelessness through ML-Informed Distribution of Rental Assistance
Catalina Vajiac, Arun Frey, Joachim Baumann, Abigail Smith, Kasun Amarasinghe, Alice Lai, Kit Rodolfa, Rayid Ghani
AAAI 2024

Previous experience

I previously worked as a Machine Learning Engineer at 3M on digital assistants for the medical domain and as an Applied Scientist at Microsoft on writing assistance tools for Word with Chris Quirk. My work in industry has included collecting data to fine-tune large transformer models, applying knowledge distillation to produce fast and accurate models for production, and developing compact task-specific representations to improve model efficiency.

I graduated in 2018 with a Ph.D. in Computer Science from the University of Illinois at Urbana-Champaign, advised by Julia Hockenmaier. My PhD research focused on how to tackle semantic tasks like textual entailment using image caption denotations. I also interned with Joel Tetreault at Grammarly in 2017 and with the Amazon A9 Product Search group in 2016.

PhD publications

Discourse Coherence in the Wild: A Dataset, Evaluation, and Methods
Alice Lai and Joel Tetreault
SIGDIAL 2018
[Code] [Data] [Supplementary Material]

Natural Language Inference from Multiple Premises
Alice Lai, Yonatan Bisk, and Julia Hockenmaier
IJCNLP 2017
[Code] [Data]

Learning to Predict Denotational Probabilities For Modeling Entailment
Alice Lai and Julia Hockenmaier
EACL 2017
[Code] [Data]

Illinois-LH: A Denotational and Distributional Approach to Semantics
Alice Lai and Julia Hockenmaier
SemEval 2014

From image description to visual denotations: New similarity metrics for semantic inference over event descriptions
Peter Young, Alice Lai, Micah Hodosh, and Julia Hockenmaier
TACL, Vol. 2
[Code] [Data]
NOTE: I no longer have access to the website that hosts the Flickr30K dataset. For any issues downloading the dataset, please contact Julia Hockenmaier.

Thesis

Textual Entailment from Image Caption Denotations
PhD Thesis, 2018
Advised by Julia Hockenmaier
Committee: Katrin Erk, Dan Roth, ChengXiang Zhai