情报科学 ›› 2025, Vol. 43 ›› Issue (7): 48-55.

• 理论研究 • 上一篇    下一篇

融合证素知识的中医证候术语映射研究

  

  • 出版日期:2025-07-05 发布日期:2025-10-16

  • Online:2025-07-05 Published:2025-10-16

摘要:

【目的/意义】探究中医证素与证候名称的关系,解决临床实践证候用语与标准证候术语之间的不匹配问
题。【方法/过程】借鉴文献著录对文献内容特征和外表特征进行客观描述的理念,基于中医证素知识分析中医证候
术语内容特征,运用词袋模型建模其形式特征,构建融合证素的多特征术语映射方法。【结果/结论】采用真实临床
病案数据进行实证。结果显示,目标术语的候选项数量N取10时,映射结果命中率提升至95.97%,当目标术语候选
项数量N减至1时,命中率提升7.18%。验证了证素先验知识术语映射方面的贡献,为中医证候术语的标准化和规
范化提供了新的思路和工具。【创新/局限】本文提出一种证素先验知识结合词袋模型的术语映射方法,将中医证候
术语的内容特征与形式特征融合,以弥补单一特征模型的视野局限性。未来将结合大模型,在更多数据上进行验
证探索。

Abstract:

【Purpose/significance】Explore the relationship between traditional Chinese medicine (TCM) syndrome elements and syn⁃
drome names, and resolve the mismatch between syndrome terms used in clinical practice and standard syndrome terminology.
【Method/process】Drawing on the idea of objectively describing the content and appearance characteristics of literature in literature re⁃
cords, the content characteristics of TCM syndrome terms were analyzed based on the knowledge of TCM syndrome elements, and the
bag-of-words model was used to model their formal characteristics. A multi-feature term mapping method that integrated syndrome el⁃
ements was constructed.【Result/conclusion】Empirical validation using real clinical case data. The results show that when the num⁃
ber of candidate items N of the target term is 10, the hit rate of the mapping result increases to 95.97%, and when the number of candi⁃
date items N of the target term is reduced to 1, the hit rate increases by 7.18%. This verifies the contribution of the mapping of
syndrome-element prior knowledge terms and provides new ideas and tools for the standardization and normalization of TCM syn⁃
drome terminology.【Innovation/limitation】This paper proposes a term mapping method that combines syndrome prior knowledge with
the bag-of-words model, which integrates the content and form features of TCM syndrome terms to make up for the field of vision limi⁃
tations of a single feature model. In the future, we will combine a large model to conduct verification and exploration on more data.