We develop and apply natural language processing techniques to analyze unstructured health data such as clinical notes, medical literature, health surveys, and patient feedback — unlocking insights that traditional methods often miss.
Much of our published research lives here: AfroLM, a multilingual language model covering 23 African languages; AfriMTE and AfriCOMET, evaluation tools for under-resourced machine translation; and CIRAL, a test collection for cross-lingual information retrieval. Together they extend the reach of health NLP to languages and communities that mainstream tools leave behind.
Pretrained models like AfroLM that bring NLP coverage to dozens of under-resourced languages.
AfriMTE and AfriCOMET — measuring machine translation quality where it was previously unmeasurable.
CIRAL and related work that makes health information findable across language boundaries.
Extracting structured insight from clinical notes, surveys, and patient feedback.