John M. Prager, IBM T.J. Watson Research Ctr.
When/Where? Friday, Nov. 3, 12:00-1:00 pm, Manning 01 (new room from year’s past)
Title: From Question Answering to Question Anticipation: The Trajectory of Watson in Electronic Medical Record Analysis
Abstract: In this talk, I will cover the evolution of the recent work in the IBM Watson Lab through a series of activities in the area of clinical decision support. I'll start with describing an adaptation of our question-answering system to answer medical licensing exam (USMLE) questions, progress through an intelligent search capability, to a dashboard presenting the clinician with information to support the treatment of a patient with minimal user input. Most of this work has been undertaken in the EMRA (Electronic Medical Record Analysis) project, and involves the development of several technologies for the analysis of medical text. These include extending the coverage of UMLS concepts and relations using knowledge of grammar, analogy and inference. I will end with a brief look at new work quantifying how much of an electronic medical record might actually be useful. If possible, I will give a demo of some of these applications.
Bio: John Prager has been working in technical fields related to using AI techniques to satisfy user information needs for most of his professional career. Most recently, while at the IBM T.J. Watson Research Center he has been part of efforts to adapt Watson technology to the health-care and other domains. In the EMRA project he has developed, or directed development of, applications in clinical decision support and invented components to extend NLP technology for the medical domain in general.
He was one of the original researchers on the Watson project, a system that played (and won) the Jeopardy! TV quiz-show game. He was involved in both the algorithms area, concentrating on question analysis and wordplay, and strategy. Previously, he led IBM’s successful entries in the TREC-QA tasks, an annual evaluation at NIST, and was co-PI on grants under the AQUAINT program. Prior to that, he worked in various areas of Search, including Language Identification, Web Search and Categorization. He has contributed components to the IBM Intelligent Miner for Text product. For a while in the early 90’s he ran the search service on www.ibm.com, and in so doing invented a concept that later became known as the hashtag.
While at the IBM Cambridge (Mass.) Scientific Center, John was the project leader of the REASON (Real-time Explanation And SuggestiON) project; REASON would provide users help by taking natural-language questions and processing them with an inference engine tied to a large repository of facts and rules about network-wide resources. John has degrees in Mathematics and Computer Science from the University of Cambridge (England) and in Artificial Intelligence from the University of Massachusetts; his publications include conference and journal papers, sixteen patents, several book chapters and a book on Turing.