Scholarlink

In Bulgarian healthcare there are some unique challenges, due to the nature of the medical data, being unstructured. Some key predictors for health such as the smoking status are described in an unstructured way in medical records, which makes impossible simple retrieval with or without regular expressions, or manual check due to the large number of records available. Large Language Models are a viable alternative to automate the process of extraction of medical data from unstructured records due to their ability to process natural language, even some for which they lack enough or specific training, such as Bulgarian language. We develop a method and a procedure to test multiple LLMs for that purpose and obtain some promising results that show their capabilities. We analyze the results in the context of the price of the service and the computational costs.

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

Large Language Models for Identification of Medical Data in Unstructured Records

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Presian Petkov,
  • Latchezar Tomov,
  • Emanuil Markov

摘要

In Bulgarian healthcare there are some unique challenges, due to the nature of the medical data, being unstructured. Some key predictors for health such as the smoking status are described in an unstructured way in medical records, which makes impossible simple retrieval with or without regular expressions, or manual check due to the large number of records available. Large Language Models are a viable alternative to automate the process of extraction of medical data from unstructured records due to their ability to process natural language, even some for which they lack enough or specific training, such as Bulgarian language. We develop a method and a procedure to test multiple LLMs for that purpose and obtain some promising results that show their capabilities. We analyze the results in the context of the price of the service and the computational costs.