Promptdiff: a fixed-point algorithm for comparing ontology versions

TLDR

Ontology development is increasingly collaborative, and unlike software code, ontologies can be identical yet differ in text, making standard text‑based version comparison ineffective. The authors created PROMPTDIFF, an algorithm that integrates heuristic matchers to compare ontology versions. PROMPTDIFF reuses software‑versioning mechanisms by iteratively applying up to ten heuristic matchers in a fixed‑point manner until no further changes occur, and the approach is extensible to more matchers. Evaluation on large projects showed PROMPTDIFF correctly identified 96 % of matches between ontology versions.

Abstract

As ontology development becomes a more ubiquitous and collaborative process, the developers face the problem of maintaining versions of ontologies akin to maintaining versions of software code in large software projects. Versioning systems for software code provide mechanisms for tracking versions, checking out versions for editing, comparing different versions, and so on. We can directly reuse many of these mechanisms for ontology versioning. However, version comparison for code is based on comparing text files--an approach that does not work for comparing ontologies. Two ontologies can be identical but have different text representation. We have developed the PROMPTDIFF algorithm, which integrates different heuristic matchers for comparing ontology versions. We combine these matchers in a fixed-point manner, using the results of one matcher as an input for others until the matchers produce no more changes. The current implementation includes ten matchers but the approach is easily extendable to an arbitrary number of matchers. Our evaluation showed that PROMPTDIFF correctly identified 96% of the matches in ontology versions from large projects.