Scholarlink

<p>The exponential growth of biomedical literature, clinical records, and health-related publications necessitates robust and scalable methods for automatic text summarization. Recent advancements in large language models (LLMs), such as BioBERT, PubMedBERT, and Med-PaLM, have significantly enhanced the capabilities of natural language processing systems to generate concise, coherent, and contextually relevant summaries tailored to the biomedical domain. This survey provides a comprehensive review of state-of-the-art methodologies for biomedical text summarization using LLMs, encompassing both extractive and abstractive approaches. We systematically examine model architectures, training paradigms, domain-specific adaptations, and evaluation techniques across various biomedical subdomains. The review adheres to the PRISMA methodology, ensuring a rigorous and reproducible literature selection process from 2016 to 2024, drawing from reputed databases including PubMed, Scopus, IEEE Xplore, and ACM Digital Library. In addition to synthesizing recent advances, the paper critically outlines prevailing challenges that impede the deployment of LLMs in biomedical summarization tasks. These include the complexity of domain-specific vocabulary and abbreviations, risks of hallucination and factual inconsistency, limited availability of high-quality annotated datasets, inadequacy of standard evaluation metrics in capturing clinical relevance, and ethical concerns surrounding data privacy and bias. Addressing these challenges is imperative to advancing the development of accurate, reliable, and ethically sound summarization systems for biomedical applications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Biomedical text summarization with large language models: methodologies, challenges, and future directions

  • 期刊论文

  • 发表日期：

    2026年1月12日

查看全文
  • Shafiya Mushtaq,
  • K. Veningston

摘要

The exponential growth of biomedical literature, clinical records, and health-related publications necessitates robust and scalable methods for automatic text summarization. Recent advancements in large language models (LLMs), such as BioBERT, PubMedBERT, and Med-PaLM, have significantly enhanced the capabilities of natural language processing systems to generate concise, coherent, and contextually relevant summaries tailored to the biomedical domain. This survey provides a comprehensive review of state-of-the-art methodologies for biomedical text summarization using LLMs, encompassing both extractive and abstractive approaches. We systematically examine model architectures, training paradigms, domain-specific adaptations, and evaluation techniques across various biomedical subdomains. The review adheres to the PRISMA methodology, ensuring a rigorous and reproducible literature selection process from 2016 to 2024, drawing from reputed databases including PubMed, Scopus, IEEE Xplore, and ACM Digital Library. In addition to synthesizing recent advances, the paper critically outlines prevailing challenges that impede the deployment of LLMs in biomedical summarization tasks. These include the complexity of domain-specific vocabulary and abbreviations, risks of hallucination and factual inconsistency, limited availability of high-quality annotated datasets, inadequacy of standard evaluation metrics in capturing clinical relevance, and ethical concerns surrounding data privacy and bias. Addressing these challenges is imperative to advancing the development of accurate, reliable, and ethically sound summarization systems for biomedical applications.