Research On Approaches To Handling Logical Conflict In OWL DL Ontologies

With continuous development of World Wide Web, the increasing Web information brings people abun-dant resources. At the same time, this results in that people cannot obtain precisely the content they want in an automatic way. To resolve this problem, the Semantic Web was proposed to better enable computers and people to work in cooperation. So far, the Semantic Web has gained great development. In the Semantic Web, ontologies as formal knowledge representation provide a shared vocabulary for different domains and facili-tate computers to deal with Web information. Web Ontology Language (OWL), based on Description Logics, is recommended by international World Wide Web Consortium (W3C). OWL provides logical foundation for the Semantic Web, especially its sublanguge OWL DL, which possesses the reasoning mechanism that can be terminated within limited time. Through reasoning with OWL DL ontologies, implicit knowledge could be inferred from those explicitly declared knowledge. However, during the processes like ontology construction, evolution and mapping, logical conflicts always exist which consists of inconsistency and incoherence. Since meaningless conclusions can be obtained when reasoning with inconsistent ontologies by a standard reasoner and incoherence is a main potential factor to cause inconsistency, it is urgent to deal with logical conflicts. The main method to deal with logical conflicts includes ontology debugging and ontology repair. However, the efficiency of existing ontology debugging approaches is relatively low. The ontology debugging systems lack the measures to assess effectiveness and dataset containing various ontologies. Besides, ontology repair approaches seldom utilize the partial order structure in ontologies. To address these problems, this paper provides the following solutions.1) With respect to ontology debugging approaches, a novel efficient ontology debugging approach based on patterns is proposed. For each selected pattern, this approach provides an algorithm to instantiate the pattern and the proof of its soundness. The experimental results show that, this approach can find more explanations in many cases and its efficiency is much higher than the selected glass-box algorithm. Besides, understanding these explanations becomes much easier.2) With respect to ontology debugging systems, the formal definitions of measuring the effectiveness of an ontology debugging system is first given. Besides, a dataset containing various ontologies is constructed and a complete evaluation for existing ontology debugging systems is first performed. The experimental results show that, existing ontology debugging systems have their advantages and disadvantages regarding the correctness, completeness and efficiency. In addition, the users can select an ontology debugging system according to different dimensions.3) With respect to ontology repair approaches, a theoretical framework to resolve conflicts in a partially or-dered ontology is first provided. Through partitioning and reordering conflicts, this framework computes the prioritized hitting sets that are as minimal as possible. Based on this framework, two new algorithms to resolve conflicts in a partially ordered ontology are proposed. The experimental results show that, partitioning conflicts can improve the efficiency of ontology repair and reordering conflicts can reduce the redundancy of prioritized hitting sets.