Matches in ScholarlyData for { ?s ?p When different versions of an ontology are published online, the links between them are often lost as the standard mechanisms (such as owl:versionInfo and owl:priorVersion) to expose these links are rarely used. This generates issues in scenarios where people or applications are required to make use of large scale, heterogenous ontology collections, implicitly containing multiple versions of ontologies. In this paper, we propose a method to detect automatically versioning links between ontologies which are available online through a Semantic Web search engine. Our approach is based on two main steps. The first step selects candidate pairs of ontologies by using versioning information expressed in their identifiers. In the second step, these candidate pairs are characterized through a set of features, including similarity measures, and classified by using Machine Learning Techniques, to distinguish the pairs that represent versions from the ones that do not. We discuss the features used, the methodology employed to train the classifiers and the precision obtained when applying this approach on the collection of ontologies of the Watson Semantic Web search engine.. }
Showing items 1 to 1 of
1
with 100 items per page.
- 56 abstract "When different versions of an ontology are published online, the links between them are often lost as the standard mechanisms (such as owl:versionInfo and owl:priorVersion) to expose these links are rarely used. This generates issues in scenarios where people or applications are required to make use of large scale, heterogenous ontology collections, implicitly containing multiple versions of ontologies. In this paper, we propose a method to detect automatically versioning links between ontologies which are available online through a Semantic Web search engine. Our approach is based on two main steps. The first step selects candidate pairs of ontologies by using versioning information expressed in their identifiers. In the second step, these candidate pairs are characterized through a set of features, including similarity measures, and classified by using Machine Learning Techniques, to distinguish the pairs that represent versions from the ones that do not. We discuss the features used, the methodology employed to train the classifiers and the precision obtained when applying this approach on the collection of ontologies of the Watson Semantic Web search engine.".