Scholarlink

<p>Medical text classification plays a vital role in intelligent diagnosis and treatment systems. In particular, sentence-level short text classification tasks are crucial for extracting key information from clinical notes, patient feedback, and medical literature, thereby assisting physicians in clinical decision-making. Chinese medical short texts have sparse features, uneven semantic distribution, and strong contextual dependency. As a result, existing classification methods often struggle to capture essential details, leading to suboptimal performance. To address this issue, we propose a Multi-scale Feature Fusion Module (MsFFM-PLM) based on RoBERTa-wwm-ext, specifically designed for text classification tasks in medical scenarios. MsFFM-PLM integrates fine-tuning strategies with a multi-scale feature fusion mechanism, while also leveraging large-scale clinical corpora to further train the pre-trained language model, jointly enhancing model performance. Experiments on benchmark datasets demonstrate that MsFFM-PLM outperforms existing mainstream models across multiple evaluation metrics.</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

A multi-scale feature fusion module based on pre-training language models for medical text classification

  • 期刊论文

  • 发表日期：

    2026年3月10日

查看全文
  • Xiaoxue Tao,
  • Mingchuan Zhang,
  • Lin Wang,
  • Youming Ge,
  • Beibei Han

摘要

Medical text classification plays a vital role in intelligent diagnosis and treatment systems. In particular, sentence-level short text classification tasks are crucial for extracting key information from clinical notes, patient feedback, and medical literature, thereby assisting physicians in clinical decision-making. Chinese medical short texts have sparse features, uneven semantic distribution, and strong contextual dependency. As a result, existing classification methods often struggle to capture essential details, leading to suboptimal performance. To address this issue, we propose a Multi-scale Feature Fusion Module (MsFFM-PLM) based on RoBERTa-wwm-ext, specifically designed for text classification tasks in medical scenarios. MsFFM-PLM integrates fine-tuning strategies with a multi-scale feature fusion mechanism, while also leveraging large-scale clinical corpora to further train the pre-trained language model, jointly enhancing model performance. Experiments on benchmark datasets demonstrate that MsFFM-PLM outperforms existing mainstream models across multiple evaluation metrics.