Informatics Research Seminar: Natural Language Processing of Medical Guidelines
January 29 @ 4:00 pm - 5:00 pm EST
Speaker: Wlodek Zadrozny, PhD
Presenting University: UNC-Charlotte; Presented from Duke University where Dr. Zadrozny is currently on sabbatical.
Broadcast Link: Seminar
There are over 30 thousand medical treatments guidelines on PubMed. However, not much is known about the properties of this corpus. For example, we do not know how often and to what degree guidelines (for the same condition) disagree.
This seminar will discuss challenges of building programs extracting information from the text of medical guidelines. A few case studies will be presented, including (1) extraction of condition-action sequences, (2) contradictions and disagreements in guidelines for the same condition. Work shown will include computational architecture for solving some of the underlying technical problems, including a searchable repository currently being built. (This is joint work with H.Hematialam and L.Garbayo, and was partly supported by IBM Faculty Research Grant).
Wlodek Zadrozny, PhD is a visiting professor with the Duke data science program (MIDS), and Professor of Computer Science at UNC Charlotte. Until 2013 he was a researcher and manager at IBM T.J. Watson Research Lab in Yorktown Heights NY. Dr. Zadrozny has about 30-years experience in natural language processing, and one of the highlights of his career was the Jeopardy! winning program IBM Watson. His research focuses on computational models of technical text, including text of patents and medical treatment guidelines.