Jay Urbain, Ph.D.
Adjunct Professor Milwaukee School of Engineering
- Milwaukee WI
Dr. Jay Urbain is an expert in machine learning, data science, and high-performance database systems.
Milwaukee School of Engineering
View more experts managed by Milwaukee School of Engineering
Education, Licensure and Certification
Ph.D.
Computer Science
Illinois Institute of Technology
2008
MBA
Business
University of Wisconsin-Madison
1997
MSEE
Computer Engineering
Illinois Institute of Technology
1989
B.S.
Engineering
Illinois Institute of Technology
1981
B.S.
Biology
Northeastern Illinois University
1979
Areas of Expertise
Accomplishments
Karl O. Werwath Engineering Research Award, MSOE
2013
Social
Research Grants
Identifying Opioid Response Phenotypes in Low Back Pain Electronic Health Data
NIH Career Grant Mentor for Dr. Meredith Adams
2017-2020
Patient Centric Medical Records Translation
NIH CTSI Pilot Award Grant
April 2018-2019, joint-PI
Honeywell, Big Data: Machine Learning of Structured and Unstructured Data
National Nuclear Security Administration
2017, Principal investigator
Honeywell, Data Modeling and Predictive Analytics of Sensor Data
National Nuclear Security Administration
2016, Principal investigator
Participated in the Biomedical Informatics Working Group, and Compiled the Biomedical Informatics Section of the Grant
NIH CTSA Grant, CTSI of SE Wisconsin
2015-2018, Co-lead, Biomedical Informatics
Selected Publications
Natural interfaces for evaluation and management of shoulder dysfunction
Innovation in AgingKorducki, J., Tidemann, A., Tarima, S., Grindel, S., Urbain, J., Mickschl, D., Rosenthal, A., Burns, E.
2018-11-01
Shoulder dysfunction affects >50% of individuals >/=60 years. Physical therapy (PT) is an effective and standard treatment component, yet adherence is low and knowledge on home exercise adherence is sparse. We developed an exploratory study to determine whether supplementing PT with objective visual feedback increases adherence to PT and home exercise (HE). Novel software was developed for a motion-sensing Kinect camera, translating video images of patients performing standard range of motion (ROM) maneuvers into skeletal avatars. Images were recorded, stored, and played back, providing objective visual documentation of progress. Participants were randomized to the intervention group (IG, viewed current and previous images a teach of 4 study visits) or control group (CG, imaged but did not view clips). All participants completed questionnaires regarding daily function, perceived improvement, PT effectiveness, and time spent on HE. 21 patients mean age 66.8 + 6.5 years attending physical therapy for first-time management of shoulder impairment were consented and enrolled. IG and CG were similar in age, demographics, and dysfunction due to shoulder impairment. IG spent more time performing HE, 42.1+/-29.8 min/day vs. 19.1+/-8.2min/day (p = 0.04, 2-sample t test). IG experienced 0% study dropout rate vs. 40% for CG (Fisher exact test; p=0.04). Adjusting for baseline ROM, IG demonstrated a trend towards greater gain in forward elevation ROM from visit to visit, compared to CG (9.25o vs.0.49o, p=0.17). Providing objective visual feedback on progress was associated with greater adherence to HE, lower attrition rates from PT, and possibly greater increases in ROM recovery.
Distributional semantic concept models for entity relation discovery
Proceedings of the 1st Workshop on Vector Space Modeling for Natural Language ProcessingUrbain, J., Bushee, G., Kowalski, G.
2015-06-01
We present an ad hoc concept modeling approach using distributional semantic models to identify fine-grained entities and their relations in an online search setting. Concepts are generated from user-defined seed terms, distributional evidence, and a relational model over concept distributions. A dimensional indexing model is used for efficient aggregation of distributional, syntactic, and relational evidence. The proposed semi-supervised model allows concepts to be defined and related at varying levels of granularity and scope. Qualitative evaluations on medical records, intelligence documents, and open domain web data demonstrate the efficacy of our approach.
User-driven relational models for entity-relation search and extraction
JIWES '12 Proceedings of the 1st Joint International Workshop on Entity-Oriented and Semantic SearchUrbain, J.
2012-06-01
The ability to extract new knowledge from large datasets is one of the most significant challenges facing society. The problem spans across domains from intelligence analysis and scientific research to basic web search. Current information extraction and retrieval tools either lack the flexibility to adapt to evolving information needs or require users to sift through search results and piece together relevant information. With so much data compounded by the criticality of finding relevant information, new tools and methods are needed to discover and relate relevant pieces of information in ever expanding repositories of data.
Exploring contextual models in chemical patent search
Information Retrieval Facility ConferenceUrbain, J., Frieder, O.
2010-06-01
We explore the development of probabilistic retrieval models for integrating term statistics with entity search using multiple levels of document context to improve the performance of chemical patent search. A distributed indexing model was developed to enable efficient named entity search and aggregation of term statistics at multiple levels of patent structure including individual words, sentences, claims, descriptions, abstracts, and titles. The system can be scaled to an arbitrary number of compute instances in a cloud computing environment to support concurrent indexing and query processing operations on large patent collections.
A dimensional retrieval model for integrating semantics and statistical evidence in context for genomics literature search
Computers in Biology and MedicineUrbain, J., Goharian, N., Frieder, O.
2009-06-01
We present a dimensional information retrieval model for combining concept-based semantics and term statistics within multiple levels of document context to identify concise, variable length passages of text that answer a user query. Our results demonstrate improved search results in the presence of varying levels of semantic evidence, and higher performance using retrieval functions that combine document, as well as sentence and passage level information. Experimental results are promising. When ranking documents based on the most relevant extracted passages, the results exceed the state-of-the-art by 15.28% as assessed by the TREC 2005 Genomics track collection of 4.5 million MEDLINE citations.
