A Lexical-based Formal Concept Analysis Method to Identify Missing Concepts in the NCI Thesaurus

Area of Science:

• Biomedical Informatics
• Computational Linguistics
• Knowledge Representation

Background:

• Biomedical terminologies are crucial for data management and semantic interoperability in research.
• Existing methods for concept enrichment often rely on external knowledge sources.
• Continuous evolution of biomedical knowledge necessitates methods for identifying missing concepts.

Purpose of the Study:

• To introduce a novel lexical method for identifying potentially missing concepts in biomedical terminologies.
• To leverage intrinsic knowledge within terminologies, specifically concept names, for concept enrichment.
• To evaluate the method's effectiveness using the National Cancer Institute (NCI) Thesaurus.

Main Methods:

• Developed a Formal Concept Analysis (FCA) based lexical method.
• Constructed an FCA formal context using lexical features of concepts.
• Applied multistage intersection to formalize new concepts and detect missing ones.

Main Results:

• Applied the method to the NCI Thesaurus (Disease or Disorder sub-hierarchy).
• Identified 8,983 potentially missing concepts.
• Preliminary evaluation found 592 of these concepts in the Unified Medical Language System (UMLS).

Conclusions:

• The FCA-based lexical method effectively identifies potentially missing concepts within biomedical terminologies.
• The method utilizes a terminology's internal structure, offering an alternative to external knowledge import.
• Further validation confirmed the presence of a subset of identified concepts in the UMLS, supporting the method's utility.