Kirk Roberts
Associate Professor
The University of Texas Health Science Center at Houston
School of Biomedical Informatics
Department of Health Data Science & Artificial Intelligence
I am a methodologist in natural language processing (NLP) for biomedicine. Specifically, this means I see to advance the methods developed for biomedical NLP. I have published almost 200 scientific articles and am an internationally-recognized leader in my specialty.
This crosses a wide variety of data types, NLP tasks, and applications:
- Data: I have performed research on the following NLP data types: clinical notes in electronic health records (EHRs), biomedical literature articles (e.g., from PubMed), clinical trial descriptions (ClinicalTrials.gov), FDA drug labels, health-related social media, authoritative medical websites, and health-related conversations. This covers almost the complete range of data that biomedical NLP researchers cover.
- Tasks: I have performed research on nearly every NLP task imaginable: syntax (e.g., syntax parsing), semantics (named entity recognition, relation extraction, event recognition, semantic role labeling), discourse (co-reference, document structure, temporal/spatial processing), pragmatics (reasoning, intent), and generation (summarization, question answering).
- Applications: I have published in a wide variety of biomedical areas: genomics, oncology, immunology and infectious diseases, radiology, pathology, ophthalmology, cardiology, psychiatry, endocrinology, dentistry, and behavioral science. This has come in the form of information extraction, question answering, and chatbot-based applications.
In short, my research spans the full range of possibilities in biomedical NLP and capable of collaborating on language-related projects with any kind of biomedical expert.
My work uses the state-of-the-art methods, which at the present moment involves the use of large language models (LLMs). I have always been at the forefront of the use of machine learning (ML) for NLP. For instance, in 2025 I published the first of vision-language models (VLMs) for clinical NLP, and in 2019 I published the first use of BERT for clinical NLP. I have had the top-performing system in many community shared task competitions. In short, it is my constant goal to push the frontier of biomedical NLP methods.
School of Biomedical Informatics Faculty
Education & Training
PhD, The University of Texas at Dallas, 2013