Your browser version is outdated. We recommend that you update your browser to the latest version.

Research

This page links to my current and completed research projects. I want to thank every collaborator and funder associated with each project. 

Frailty Decision Making (ongoing): Designing an ontology for frailty as it relates to making decisions surrounding cardiac interventions. Investigating the decision making process to determine the possible underlying knowledge representation.

Sentiment in Patient Feedback: Classifying patient feedback into categories beyond the standard positive/negative.

Ontology: As part of the VA ProWATCH team I constructed an ontology of medically unexplained syndromes.

POET2/FairCode Project: High Performance Computing for Advanced Clinical Narrative Processing 

APCD: formative and summative evaluation of the Utah Department of Health Enhancing Utah All Payer Claims Database for Healthcare Cost Transparency Cycle III grant

SEAM Project: The Semi-Automated Ontology Development System 

CHV Project: Sites to facilitate access and maintenance for the Open Access Collaborative - Consumer Health Vocabulary 

 

NLP Research Interests

It apparent to me that natural language processing systems are experiencing a significant improvement in performance through the combination of Machine Learning techniques with the architecture of the human cognitive system as evidenced by the success of deep learning techniques. We created a system that did not use deep learning, but was motivated by the current understanding of human text processing. POET is a multi-level, multi-path system built in UIMA modules (e.g., dependency parsing, rule-based, ML) that reflect findings from Cognitive Science (e.g., dorsal-ventral language processing streams, differential processes indicated by the N400 ERP wave).

 

Health Care Research Interests

It is my belief that health care (diagnostic accuracy and efficiency) can be improved by accessing the wealth of information available in the free-text of electronic health records. The key is to recover and analyze the information for timely and appropriate presentation to facilitate clinicians’ cognitive processing.

Biomedical Natural Language Understanding

My research in this area focuses on designing knowledge representations will be designed to incorporate ontology grounded in realist principles determining to facilitate machine-based reasoning. This work requires developing ontologies, maintaining them, developing NLP systems that can extract the information which the knowledge representation will contain, and using that representation to reason about the information contained in clinical notes.

Clinicians’ information needs

In order to create an accurate knowledge representation it is necessary to understand how clinicians think about medical information. To this end, I have engaged in research designed to illuminate clinicians’ information needs as well as the differences in clinical vocabulary between clinical roles and clinical specialties. Doctors, nurses, and pharmacists do not record the same information in clinical records to develop and understanding of their entries it is important to distinguish the terms they use and how they use them. The same process is necessary between clinical specialties.

 

Data and Tool dissemination

Another important contribution to science is disseminating the data and tools created. I have created and currently main 4 websites, which make my work publicly available. The website detailing my postdoctoral research is kdh-nlp.com. It also provides access to our Semi-automated Management for Ontology Building system (kdh-nlp.org/SEAM). For my first postdoctoral project I build a Consumer Health Vocabulary Automatic Vocabulary Maintenance Site available at consumerhealthvocab.chpc.utah.edu/AutoVocabMaint and Consumer Health Vocabulary Wiki pages available at consumerhealthvocab.chpc.utah.edu/CHVwiki.