| Ontology has been widely used in many fields in recent years. However, during ontology evolution, incorporating newly received information from different sources with the onology would lead to an incoherent ontology. Therefore, there is an increasing interest in ontology debugging and incoherence handling. Nevertheless, there are many drawbacks on existing debugging and revision approaches. For example, in the actual debugging process, inference engine will be called many times while taking a lot of memory, which will become very difficult when dealing with huge amounts of data. Meanwhile, revision algorithms will lose entailed information of original ontology, and revision results may exceed original expressiveness.In order to solve above problems, we propose graph-construction rules which can transform a DL-Lite ontology to a directed graph, and then we give our graph-based ontology debugging algorithms and interactive revision algorithm. Specifically, our works are as follows.1) In this thesis, we propose construction rules that can transform a DL-Lite ontology to a directed graph and we prove the equivalence of the ontology and the graph constructed from it. Then, we provide graph-based ontology debugging algorithms that can find unsatisfiable concepts or roles, all MUPS and all MIPS. We also give theoretical foundations of these algorithms.2) A graph-based interactive ontology revision algorithm is provided. It can improve revision accuracy and minimize the loss of ontology information by adding original information that can be entailed.3) We implement our algorithms and develop an ontology debugging and revision system. We then evaluate the performance of these algorithms by carrying out a series of experiments on some adapted real ontologies. The experimental results show that the efficiency and scalability of our debugging algorithms outperform existing approaches and our revsion algorithm is superior while minimizing the loss of ontology information. |
PDF Full Text Request