Corpus-based associations provide additional morphological variants to medical terminologies

Area of Science:

• Medical Informatics
• Computational Linguistics
• Natural Language Processing

Background:

• Automated coding and indexing in medicine rely on recognizing term variants.
• Existing resources like the UMLS Specialist Lexicon provide morphological knowledge for English but lack equivalents for other languages.
• Developing methods to create morphological knowledge bases for diverse languages is crucial.

Purpose of the Study:

• To design general methods for collecting morphological knowledge for a given language.
• To propose and apply a method for discovering derived words within text corpora.
• To evaluate the method's effectiveness in identifying derived adjectives for medical terms.

Main Methods:

• Developed a general method for discovering derived words in text corpora.
• Applied the method to a French medical corpus.
• Evaluated the method by assessing its ability to suggest derived adjectives for nouns in the SNOMED nomenclature.

Main Results:

• The proposed method achieved 74% precision in suggesting derived adjectives.
• The method covered 16% of the studied nouns, exceeding current SNOMED coverage.
• An additional 76% of adjectives were suggested, significantly expanding SNOMED's adjectival equivalents.

Conclusions:

• The developed method effectively aids in constructing morphological knowledge bases.
• This approach can accelerate the expansion of term variants within controlled vocabularies.
• The findings support the use of corpus-based methods for enhancing medical terminologies in various languages.