EAPB: entropy-aware path-based metric for ontology quality.

Area of Science:

• Computer Science
• Information Theory
• Medical Informatics

Background:

• Entropy is increasingly used to measure knowledge base predictability and redundancy.
• Existing entropy methods for ontologies overlook path connectivity and assign uniform weights.
• This limits the accurate evaluation of complex knowledge structures like ontologies.

Purpose of the Study:

• To introduce an Entropy-Aware Path-Based (EAPB) metric for improved ontology quality assessment.
• To address limitations of current entropy applications by considering path and textual information.
• To provide a more reliable method for evaluating ontologies, particularly in medical informatics.

Main Methods:

• Developed the EAPB metric, integrating path information and textual content for network connectivity.
• Utilized structure-based and text-based embeddings multiplied by the entropy computation's connectivity matrix.
• Conducted analytical evaluations against state-of-the-art criteria and empirical analysis on real-world and synthetic ontologies.

Main Results:

• The EAPB metric effectively evaluates ontology quality, demonstrating reliability through statistical information, entropy evaluation, and case studies.
• Experimental analysis confirmed the metric's capability, especially for medical informatics ontologies.
• The approach leverages path and textual information to enrich network representational learning for entropy computation.

Conclusions:

• The EAPB metric offers a significant advancement in ontology evaluation by incorporating structural and textual data.
• It benefits semantic web analytics and management of ontology development projects.
• Results are reproducible, with source code and ontology to be released post-publication.