Benchmarking Ontologies: Bigger or Better?
Author Information
Author(s): Yao Lixia, Divoli Anna, Mayzus Ilya, Evans James A., Rzhetsky Andrey
Primary Institution: Columbia University and University of Chicago
Hypothesis
How can we assess the fidelity of an ontology's domain representation?
Conclusion
The study introduces metrics that evaluate the breadth and depth of ontologies, showing that medical ontologies and English thesauri have limited overlap in concepts and relations.
Supporting Evidence
- Medical ontologies have a small intersection, as do English thesauri.
- Different domains use the same words quite differently.
- The proposed metrics can guide improvements in ontology relevance.
Takeaway
This study looks at how well different systems of knowledge, called ontologies, represent information in medicine and language. It finds that they often don't overlap much, which means they might not be as useful as they could be.
Methodology
The study analyzes four medical ontologies and seven English thesauri against various corpora using metrics for breadth and depth of representation.
Potential Biases
The evaluation may be confounded by imperfections in text analysis and mapping technologies.
Limitations
The approach relies on the availability of large corpora related to the domain of interest, which may not be feasible for all ontologies.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website