Research On Large-scale Ontology Partition And Mapping Based On Fragment

With the quickly development of Semantic Web, ontologies become more numerous in recent years. And the scale of the ontologies also becomes larger than before. So, the ontology integration, reuse, maintenance and sharing become urgent problems. The thesis focuses on research on large ontology partitioning and mapping.Firstly, the thesis simply introduces the research background of the topic, summarizes some common ontology matching methods and systems, and the large scale ontology mapping problems are also simply introduced and analyzed.Secondly, aiming at solving the problem of the traditional method does not fully uses the semantic information of the ontologies and other problems, the new semantic similarity computation method based on ontology feature is proposed. It first expresses each concept as a set based on its feature information, the elements of set is called feature attribute, then obtains the concept similarity through calculating the similarity between two sets, then introduces the influencing factors such as width and depth to get the final similarity metric. Theoretical analysis and experimental result show that the metric is simple and has accurate results.Thirdly, in order to solve the problem of current large scale ontology mapping method is not automatic and the ontology blocks are variable in size, the thesis propose a method for large ontology partition and mapping based on segment matching. It uses the idea of dividing and conquering, then the whole problem is divided into three parts, which are ontology partition, segment matching and mapping found. First, an agglomerative hierarchy clustering algorithm is used to partition the whole large ontology into a set of ontology segment. Then use a heuristic method based on anchor to found segment mapping through analyzing the distribution of anchors. At last, it uses muti-strategies to find elements mapping between source ontology and target ontology. Finally, based on the research above, segment based larger ontology mapping (SLOM) has been designed and implemented. Experiment result shows that it is better than current system on partition quality and segment mapping quality, and this system improves the precision while maintaining recall, thus gets a better mapping result.