JuFiT: A Configurable Rule Engine for Filtering and Generating New Multilingual Umls Terms

Area of Science:

• Medical Informatics
• Natural Language Processing
• Computational Linguistics

Background:

• The Unified Medical Language System (UMLS) is a crucial resource for biomedical text processing.
• Existing tools for UMLS term extraction and integration have limitations, particularly in multilingual contexts.
• Filtering non-natural terms and enriching the UMLS are essential for improving biomedical information retrieval.

Purpose of the Study:

• To introduce JuFiT, an adjustable rule engine for processing the UMLS.
• To enable the filtering of non-natural terms and the addition of new terms to the UMLS.
• To develop a multilingual solution for UMLS term management, supporting English, Spanish, French, German, and Dutch.

Main Methods:

• JuFiT employs an easily adjustable rule engine.
• It identifies and filters terms not typically found in running citation texts.
• The system can rewrite non-natural terms to integrate them into the UMLS.

Main Results:

• JuFiT demonstrates multilingual capabilities across five prominent European languages.
• Evaluation under various experimental conditions indicates increased annotation quality for English.
• Evidence suggests potential improvements in annotation quality for German and Spanish.

Conclusions:

• JuFiT offers a flexible and multilingual approach to enhancing the UMLS.
• The tool addresses limitations of previous methods like MetaMap and Casper in multilingual settings.
• JuFiT shows promise for improving the accuracy and coverage of biomedical text annotation across different languages.