Scholarlink

Access to precise and structured medical information is essential to support clinical decision-making, foster research development, and inform health policy design. However, much of this information is available in unstructured formats, making its analysis and utilization challenging, particularly in Spanish, where technological resources are limited. This study, uniquely leveraging the CT-EBM-SP corpus (Clinical Trials for Evidence-based Medicine in Spanish), aims to identify key medical concepts in Spanish medical texts. By applying advanced Natural Language Processing (NLP) techniques and machine learning, the study provides semantic structure to medical information. The proposed approach allows for the classification of relevant medical concepts, including chemical substances (CHEM), procedures (PROC), diseases (DISO), and anatomical structures (ANAT), thus generating semantically structured clinical information. The evaluation of the proposed approach employs metrics well-established in the state of the art, such as accuracy and F1-score, to compare various feature engineering methods and algorithms. The best results were achieved using the Support Vector Machines (SVM) algorithm combined with morphological and semantic features, attaining an accuracy of 56.35%. These findings contribute to automating and optimizing information extraction in Spanish medical literature.

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

Machine Learning Based Approach for Medical Concepts Identification from Spanish Texts

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Samyllerick Tapia Hernández,
  • José A. Reyes-Ortiz,
  • Maricela Bravo

摘要

Access to precise and structured medical information is essential to support clinical decision-making, foster research development, and inform health policy design. However, much of this information is available in unstructured formats, making its analysis and utilization challenging, particularly in Spanish, where technological resources are limited. This study, uniquely leveraging the CT-EBM-SP corpus (Clinical Trials for Evidence-based Medicine in Spanish), aims to identify key medical concepts in Spanish medical texts. By applying advanced Natural Language Processing (NLP) techniques and machine learning, the study provides semantic structure to medical information. The proposed approach allows for the classification of relevant medical concepts, including chemical substances (CHEM), procedures (PROC), diseases (DISO), and anatomical structures (ANAT), thus generating semantically structured clinical information. The evaluation of the proposed approach employs metrics well-established in the state of the art, such as accuracy and F1-score, to compare various feature engineering methods and algorithms. The best results were achieved using the Support Vector Machines (SVM) algorithm combined with morphological and semantic features, attaining an accuracy of 56.35%. These findings contribute to automating and optimizing information extraction in Spanish medical literature.