Improving entity recognition using ensembles of deep learning and fine-tuned large language models: A case study on adverse event extraction from VAERS and social media.

Journal: Journal of biomedical informatics
PMID:

Abstract

OBJECTIVE: Adverse event (AE) extraction following COVID-19 vaccines from text data is crucial for monitoring and analyzing the safety profiles of immunizations, identifying potential risks and ensuring the safe use of these products. Traditional deep learning models are adept at learning intricate feature representations and dependencies in sequential data, but often require extensive labeled data. In contrast, large language models (LLMs) excel in understanding contextual information, but exhibit unstable performance on named entity recognition (NER) tasks, possibly due to their broad but unspecific training. This study aims to evaluate the effectiveness of LLMs and traditional deep learning models in AE extraction, and to assess the impact of ensembling these models on performance.

Authors

  • Yiming Li
    Department of Cardiology, West China Hospital, Sichuan University, Chengdu 610041, China.
  • Deepthi Viswaroopan
    The University of Texas Health Science Center at Houston.
  • William He
    Duke University.
  • Jianfu Li
    Mayo Clinic.
  • Xu Zuo
    The University of Texas Health Science Center at Houston.
  • Hua Xu
    Department of Urology, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China.
  • Cui Tao
    The University of Texas Health Science Center at Houston, USA.