Research On Repairing Of Ontology Mapping Based On Reasoning And Extraction Algorithm

The existing strategies of ontology alignment have their inherent drawbacks. On one hand, the thresholds can't be computed dynamically for each strategy; on the other hand, incorrect mappings are yield easily. This thesis focuses on these issues.Firstly, this thesis introduces the background of our research simply, and presents the current technique of ontology alignment of multi-strategy, the problems of each strategy threshold computation and the repairing of incorrect mappings are also analyzed.To address the problem of different contributions to similarity computation for each strategy, we design an adaptive algorithm for computing the threshold of each strategy by parsing the ontologies to be aligned, in order to take the most of semantic information for each strategy. In this thesis, two methods are given to integrate the similarity, one is to define the harmony of ontology for the combination of multi-strategy, and the other is to employ a fast-recommendation method to detect the real semantic information of the ontologies to be aligned, so give different thresholds for each strategy, respectively.Aiming at the problem of the final output similarity matrix often involves incorrect mappings when using multi-strategy, sufferage algorithm to extract the mappings based on Distribute Description Logic is proposed, which combines the technique of reasoning. By establishing repairing rules to remove the mappings which produce semantic conflicts or structure conflicts, this algorithm can improve the quality of extraction correctly and infectively.Finally, the adaptive algorithm for computing the thresholds of multi-strategy and the mapping extraction algorithm are designed and implemented. The experiment shows that our algorithms can improve the precision while maintaining recall, thus get a better mapping result.