Bio-Epidemiology-NER

Bio-Epidemiology-NER applies a Transformer-based model to perform biomedical named entity recognition on unstructured clinical and epidemiological text to extract clinical, biomedical, and sociodemographic entities for downstream analysis.

Key Features:

• Transformer-based architecture: Uses a Transformer-based model for named entity recognition.
• Comprehensive entity recognition: Identifies medical risk factors, vital signs, drugs, biological functions, age, gender, race, and social history from text.
• Training dataset: Trained on a richly annotated dataset encompassing medical, clinical, biomedical, and epidemiological named entities.
• Pre-processing and data parsing: Includes text pre-processing and data parsing pipelines to prepare and organize input text.
• Named entity enhancement: Applies refinement of recognized entities to improve accuracy and relevance.
• Scalability and configurability: Architecture supports scaling for both training and inference and configurable adaptation to different datasets.
• Performance: Reports macro- and micro-average F1 scores around 90% or higher on benchmark datasets.

Scientific Applications:

• Epidemiological and clinical research: Extracts clinical and non-clinical factors from unstructured biomedical text to support studies of health outcomes.
• Large-scale biomedical text processing: Processes large volumes of unstructured biomedical and epidemiological text for analysis at scale.
• Sociodemographic factor extraction: Identifies age, gender, race, and social history to support analyses of demographic influences on health.
• Academic research and clinical applications: Provides entity-level annotations for downstream analyses in academic studies and clinical research.

Methodology:

Pre-processing, data parsing, Transformer-based named entity recognition, and named entity enhancement (refinement).