Research

The majority of my work has been in Clinical Natural Language Processing. The overarching theme of my research under Dr. Chitta Baral has been to leverage NLP across biomedical & clinical data (like Electronic Medical Records) to make actionable decisions that improve efficiency of healthcare delivery.

Below are a few of the things I've worked on. There are also more details on Google Scholar or my LinkedIn.

Multi-Perspective Biomedical Semantic Question-Answering (MS Thesis)

Given a query in natural language, search across 29 million PubMed abstracts and identify top n candidate sentences that answer the query. To better "understand" and rank candidates, a weighted "Multi-Perspective" approach, utilizing three BERT models trained on different tasks, is taken.

Short Paper | Full Thesis

../images/semanticir.png

Clinical Trial Eligibility Classification from Electronic Medical Records

From natural-language patient EMR, classify whether patient falls within or outside 13 clinical trial cohorts (i.e. alcohol abuse, drug abuse, MI within past 6 months, advanced coronary artery disease).

Part of the n2c2 2018 Challenge – ranked #1 out of 47 teams.

Paper

../images/clinicaltrial.png

Prescription Information Extraction from Electronic Health Records (BS Thesis)

Bidirectional LSTM + CRF neural architecture for Named Entity Recognition applied to the i2b2 2009 Medication Information extraction challenge.

Undergraduate Honors Thesis

../images/medner_example.png