Active User

current user.

Data, Text, and Rule Mining

Primary area of research is data and text mining. In text mining, we often encounter situations where we have to parse millions of sentences into appropriate grammatical structures. Other situations include identifying named entities (drugs, diseases, people, etc) and analyzing sentiment expressed in textual reviews or statements. I also work in rule mining where we find statistically significant rules from large datasets.

GPUs

Primary area of research is data and text mining. Current research projects includes computational code extraction (multi-label classification), biomedical relation extraction and word sense disambiguation, literature based knowledge discovery, social media mining for health related topics, predictive modeling and association rule mining from large healthcare datasets.

DNNs

Deep neural networks (DNNs) are being actively used for longitudinal modeling of EMR data toward disease diagnosis and prognosis. Project(s) pertain to the applications of recent advances in DNNs specifically recurrent architectures and the faster attention-based transformer models to EMR modeling. Orhan Abar's dissertation work included depression as the target condition but we plan to extend this to opioid use disorder (OUD) and substance use disorder (SUD) prediction. The modeling is slightly different for these purposes given subject matter expertise is being used to select variables that need encoding via DNNs. The use-cases are first being explored for people who are diagnosed with ADHD or who are taking strong painkillers for chronic back pain. Orhan Abar may need to implement new architectures or extend the current ones to fit the new use-cases.

Students

Tung Tran - Added 7/14/20

Aviv Brokman

Jiho Noh

Patrick J Ward, Statistician KY Injury Prev & Research Center - Added 10/09/2020

Xuguang Ai, Added 11/18/2021

Yuhang Jiang, Added 11/18/2021

Shashank Gupta, Added 11/18/2021

External Leaner

Mehmet Gokhan Bakal

Collaborator

Orhan Abar, University osmaniye korkut ata, External 09/14/2020

Software

Our platforms are Java or Python, and typically involve open source software from academic institutions. Examples include MetaMap, SemRep (from the national library of medicine), Twitter4J, Vowpal Wabbit (Microsoft Research), NLTK, Scikit Learn, Theano etc.

Publications

None at this time.

Browser not supported