Word Root Translation of 45564 Autopsy Reports into MeSH Titles

摘要

It has become increasingly cost-effective to produce primary medical records as electronic documents. However, access to these documents by disease-concept is limited by the user's knowledge of the classification system (ICD, SNOMED, MeSH). We have developed a system of algorithmic translation between medical natural language and MeSH titles, by respelling common word roots. Using the list of 17,494 words harvested from all autopsies performed at The Johns Hopkins Hospital between May 28, 1889, and June 30, 1987, a set of 2,310 respelling formulas converted 10,200 words into 659 distinct MeSH titles. Results show that respelling formulas can be used to translate medical natural language into MeSH titles, and thus improve the access to the database.