Natalie Parde
University of Illinois Chicago, IL·Computer Science
How can computers that understand human language be made fair and trustworthy enough to help in health care? The grants describe two projects that apply natural language processing (NLP), the field that teaches computers to analyze text and speech, to health and well-being.
- Fields
Computer Science
Medicine & Clinical Research
Public Health & Epidemiology
- Topics
natural language processing
large language models
Alzheimer's disease
- Grants
- 2 known grants, through May 2028
What they work on
- Design large language model (LLM) methods that detect mild cognitive impairment and dementia from transcribed spontaneous speech, and explore whether they can detect the earliest preclinical stage of Alzheimer's disease.
- Make LLM-based Alzheimer's detection fairer and easier to interpret through new ways of selecting demonstration examples and through domain-aligned instruction tuning, and develop new methods to explain model output in health care settings.
- Validate the speech-based detection methods by checking how well their outputs line up with clinical measures of Alzheimer's disease and mild cognitive impairment, and test whether the findings carry over to languages beyond English.
- Adapt the Participation and Environment Measure (PEM), an electronic tool for planning early intervention services, so its content is culturally equivalent and relevant for racially and ethnically diverse families, including ways to voice concerns about racial climate.
- Build an adaptive conversational agent into the PEM that guides caregivers and personalizes its strategy-sharing feature, advancing research on dialogue policies, dialogue act schemes and lexicon-informed meaning representations.
- Evaluate the upgraded PEM with diverse and socially disadvantaged early intervention families to test effects on perceived service quality and parent engagement, and gather stakeholder input on barriers to long-term use.
More about the research
One spots early signs of Alzheimer's disease in the way people speak. The other builds a conversational tool that helps families of young children with developmental delays help plan their child's therapy. In both, the goal is to make the systems work equally well for diverse groups of people and to make their outputs easy to explain.
Why it matters
Speech-based screening could offer a low-cost, scalable, non-invasive way to catch Alzheimer's disease and mild cognitive impairment earlier, and fairer models could widen access to cognitive health assessment across populations. A culturally relevant, AI-guided planning tool could help racially and ethnically diverse and socially disadvantaged families of infants and toddlers with developmental delays get higher-quality, family-centered early intervention services.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Catherine Clelland
Columbia University, NY·Neuroscience
Test whether disrupted BH4 production, driven by functional variants in the GCH1 gene that controls BH4 synthesis, lowers BH4 levels and predicts cognitive impairment in patients with Alzheimer's disease and mild cognitive impairment.
Anthony Sicilia
West Virginia University, WV·Computer Science
Develop a Probably Approximately Correct (PAC) learning theory that measures how user disagreement, suggestions and topic shifts change the calibration of an AI system's confidence during a conversation.
Chirag Shah
University of Washington, WA·Computer Science
Build a proactive conversational agent on an offline hierarchical model-based deep reinforcement learning framework, in which a high level picks conversation strategies and a low level generates responses.
The school
University of Illinois Chicago: 42 labs on PI Pages, $532M a year on research, 359 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.