Deep Learning for Unstructured Healthcare Data: From Adverse Event Detection to EHR Standardization

This presentation explores the application of large language models (LLMs) and deep learning to extract actionable insights from unstructured healthcare data. The talk addresses LLM-based detection of drug discontinuation events and adverse events from social media data. Leveraging LLMs through zero-shot inference and task-specific fine-tuning, we demonstrate robust detection of medication-related events from online health communities, enabling population-level pharmacovigilance in real-world, data-sparse settings. Our analyses illustrate how social media discourse can complement traditional adverse event registries and support public health surveillance. The talk also covers our work on AI-assisted standardization of nursing care plan data from electronic health records. Nursing documentation remains largely unstandardized, limiting its integration into clinical research networks. We developed a Retrieval-Augmented Generation pipeline using LLMs to map local nursing terms to standardized terminologies, and evaluated various prompt engineering techniques to optimize mapping accuracy. This approach significantly reduces manual effort and advances semantic interoperability of nursing data.

Dr. Alexander Semenov is an Assistant Professor in the Department of Industrial and Management Systems Engineering at the College of Engineering, University of South Florida. He received his Ph.D. in Computer Science from the University of Jyväskylä, Finland. His research and teaching interests include network science, social media analytics, efficient algorithm design, large-scale data analysis, optimization, and machine learning. He has co-authored over 80 peer-reviewed publications and has been a recipient of research grants from agencies such as NIFA USDA, AFOSR, and Business Finland. Dr. Semenov is an Associate Editor of the Journal of Combinatorial Optimization, Energy Systems, and IET Blockchain journal. He is also a member of the editorial board of Scientific Reports, conferences such as S&P, CCS, USENIX Security, NDSS, SIGCOMM, NSDI, NeurIPS, ICML, and CHI, and well-recognized journals such as IEEE TIFS, IEEE TDSC, IEEE/ACM TON, and IEEE TKDE. He is the recipient of IEEE Big Data Security Senior Research Award in 2025, ACM SACMAT Test-of-Time Award in 2024, and the Best Paper Awards from ACM ASIACCS (2022), ACSAC (2020), IEEE ICC (2020), ACM SIGCSE (2018), and ACM CODASPY (2014). His research has won the First Place Award in ACM SIGCOMM 2018 SRC. His research has also been featured by the IEEE Special Technical Community on Social Networking and received 50+ press coverage including ACM TechNews, InformationWeek, Slashdot, etc.