UFOme: An ontology mapping system with strategy prediction capabilities

摘要

Ontology mapping, or matching, aims at identifying correspondences among entities in different ontologies. Several strands of research come up with algorithms often combining multiple mapping strategies to improve the mapping accuracy. However, few approaches have systematically investigated the requirements of a mapping system both from the functional (i.e., the features that are required) and user point of view (i.e., how the user can exploit these features). This paper presents an ontology mapping software framework that has been designed and implemented to help users (both expert and non-expert) in designing and/or exploiting comprehensive mapping systems. It is based on a library of mapping modules implementing functions such as discovering mappings or evaluating mapping strategies. In particular, the strategy predictor module of the designed framework, for each specific mapping task, can “predict” mapping modules to be exploited and parameter values (e.g., weights and thresholds). The implemented system, called UFOme, assists users during the various phases of a mapping task execution by providing a user friendly ontology mapping environment. The UFOme implementation and its prediction capabilities and accuracy were evaluated on the Ontology Alignment Evaluation Initiative tests with encouraging results.