Artificial intelligence driven context-aware biomedical information extraction model for multi-class disease condition classification using NLP with GPT-3

Abstract Precise extraction of medical phenotypes and entities from electronic health record (EHR) text is vital for numerous clinical research tasks, including cohort classification, tracking temporal patterns in disease evolution, and developing treatment plans. Still, this task remains difficult owing to the ambiguity and complexity of medical language. The application of generative pre-trained (GPT) techniques, such as GPT-3, GPT-4, and GPT-5, is explored for extracting biomedical information. Also, large language models (LLMs) are evaluated to improve their quality. Recently, rapid developments in artificial intelligence (AI) have significantly influenced many areas, with healthcare standing out as the most transformative. This is particularly evident in the field of natural language processing (NLP), where AI technologies capable of understanding and generating human-like text have altered how healthcare services are provided. In this manuscript, an Automated Biomedical Information Extraction Using Generative Pre-Training and Hybrid Attention Model (ABIE-GPTHAM) methodology is proposed in contextual NLP. This paper aims to develop a context-aware biomedical information extraction framework using AI to accurately identify and interpret relevant clinical entities from unstructured medical text. At first, the text pre-processing phase involves crucial steps such as tokenisation, punctuation removal, stop-word removal, case conversion, and stemming or lemmatisation to prepare the text for further analysis. Furthermore, the generative pre-training-3 (GPT-3) method is used to obtain the word vector representation. Moreover, an attention-based bidirectional long short-term memory (A-BiLSTM) method is employed for classification. The comparison analysis of the ABIE-GPTHAM approach demonstrated a superior accuracy of 98.80% compared to existing methods on the medical text dataset.