The Evolving Landscape of Generative Large Language Models and Traditional Natural Language Processing in Medicine

2025-05-15

AI Summary

Clarifying the capability boundaries and complementary roles of generative large language models (LLMs) versus traditional NLP methods in medical AI applications remains an open challenge. Method: We conducted a large-scale bibliometric analysis of 19,123 publications, systematically categorized medical tasks, and performed cross-paradigm performance benchmarking across diverse clinical NLP tasks. Contribution/Results: Our study provides the first large-scale empirical evidence that generative LLMs significantly outperform traditional methods in open-ended generative tasks—such as clinical question answering and medical report generation—while maintaining robust safety and controllability. Conversely, conventional NLP approaches retain substantial advantages in structured information extraction tasks—including named entity recognition and relation extraction—demonstrating superior accuracy and interpretability. Based on these findings, we propose a task-aware technology selection framework tailored to medical domain characteristics, offering both theoretical foundations and practical guidelines for ethically grounded, efficient, and context-appropriate deployment of AI in healthcare.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Natural language processing (NLP) has been traditionally applied to medicine, and generative large language models (LLMs) have become prominent recently. However, the differences between them across different medical tasks remain underexplored. We analyzed 19,123 studies, finding that generative LLMs demonstrate advantages in open-ended tasks, while traditional NLP dominates in information extraction and analysis tasks. As these technologies advance, ethical use of them is essential to ensure their potential in medical applications.

Problem

• Comparing generative LLMs and traditional NLP in medical tasks
• Identifying strengths of LLMs in open-ended medical tasks
• Ensuring ethical use of NLP technologies in medicine

Innovation

• Generative LLMs excel in open-ended medical tasks
• Traditional NLP leads in information extraction tasks
• Ethical use is crucial for medical applications