Abstract

We challenge existing query-based ontology fault localization methods wrt.assumptions they make, criteria they optimize, and interaction means they use.We find that their efficiency depends largely on the behavior of theinteracting expert, that performed calculations can be inefficient orimprecise, and that used optimization criteria are often not fully realistic.As a remedy, we suggest a novel (and simpler) interaction approach whichovercomes all identified problems and, in comprehensive experiments on faultyreal-world ontologies, enables a successful fault localization while requiringfewer expert interactions in 66 % of the cases, and always at least 80 % lessexpert waiting time, compared to existing methods.

 

Quick Read (beta)