Using WordNet to improve the mapping of data elements to UMLS for data sources integration

Area of Science:

• Biomedical Informatics
• Computational Linguistics

Background:

• Biomedical systems utilize distinct data elements (DEs) for information representation.
• Integrating these DEs is essential for unifying biomedical resources.
• Mapping DEs to the Unified Medical Language System (UMLS) often presents ambiguity or lacks direct correspondence.

Purpose of the Study:

• To evaluate the utility of WordNet, a general-purpose lexical database, in enhancing DEs mapping to the UMLS.
• To explore WordNet's potential in validating, disambiguating, and identifying indirect mappings for DEs within biomedical contexts.

Main Methods:

• The study employed synonyms, definitions, and structural properties from WordNet.
• A set of 474 DEs from eleven diverse biomedical sources was analyzed.
• WordNet's contribution was assessed by comparing its performance against direct UMLS mappings.

Main Results:

• WordNet demonstrated effectiveness in validating and disambiguating direct DEs-to-UMLS mappings.
• The use of WordNet facilitated the identification of indirect mappings between DEs and the UMLS.
• The proposed method showed significant potential for improving biomedical resource integration.

Conclusions:

• WordNet serves as a valuable supplementary resource for enhancing the accuracy and completeness of DEs mapping to the UMLS.
• Leveraging general terminologies like WordNet can overcome limitations in current biomedical terminological mapping.
• This approach offers a promising strategy for more robust biomedical data integration.