Algorithms to evaluate ontologies based on extended error taxonomy

Ontology evaluation is an important part of the ontology development process. Errors in ontology are dangerous for the information system built on it. To our surprise, the existing systems are unable to identify most of the errors. We could not find algorithms to detect most of the errors in the published literature. We design and implement algorithms to detect various errors in ontologies: circulatory errors in class and property hierarchy, common class and property in disjoint decomposition, redundancy of sub class and sub property, redundancy of disjoint relation and disjoint knowledge omission. Some of the ontologies are very large in size. Some warnings need evaluation by human expert. So the algorithms employ a mechanism to rank the warnings such that the warnings with high probability of realization by expert come towards the top. For the implementation, ontologies are indexed using a variant of already proposed and published scheme Ontrel. In addition to the previous errors taxonomy, the algorithms also cover recently extended error taxonomy: circulatory errors in property hierarchy, common property in disjoint decomposition, redundancy of sub property, redundancy of disjoint relation and disjoint knowledge omission. Using these algorithms, we evaluate well known ontologies including Gene Ontology (GO), WordNet Ontology, OntoSem Ontology and identify errors in all of them. Since critical systems are being developed based on ontologies, this initiative of evaluation will be a significant contribution in the ontology development process.