About

We develop methods to improve NLP tools for low-resource languages—those lacking the abundant data needed to train modern machine learning models.

Most ML approaches require vast amounts of text (hundreds of gigabytes), available only for high-resource languages like English and Chinese. This leaves the majority of the world's languages behind, undermining the vital role these systems can play in tools like keyboard autocorrect, speech recognition, and machine translation—tools that help languages thrive in the digital era.

To address this gap, we focus on machine learning techniques that work with limited data:

• Unsupervised/self-supervised learning — training with raw text or much smaller amounts of specialized data
• Multilingual modeling — pooling language data by training on multiple languages at once
• Transfer learning — leveraging models from higher-resource languages for new, low-resource ones

For a complete list of publications, see lab members' individual pages.

Selected Recent Publications

• Linguistically Informed Evaluation of Multilingual ASR for African Languages
  Fei-Yueh Chen, Lateef Adeleke, C.M. Downey
  7th Workshop on African Natural Language Processing (AfricaNLP), 2026
  Paper
• Targeted Multilingual Adaptation for Low-resource Language Families
  C.M. Downey, Terra Blevins, Dhwani Serai, Dwija Parikh, Shane Steinert-Threlkeld
  Findings of EMNLP, 2024
  Paper