Multi-strategy Ontology Mapping Based on Stacking Method
Hongke Xia, Xuefeng Zheng, Xiang Hu, Yingai Tian · 2009
As a crucial point of semantic integration, ontology mapping has been researched in-depth in recent years. In this paper, the problem of concept similarity computing is regarded as the problem of classifying, and a multi-strategy ontology mapping framework based on stacking method is proposed with the multi concept similarity computing strategies being combined with Stacking method; also, a novel metadata classifying algorithm LMSMC based on Widrow-Hoff theory is presented. In this framework, the target ontology is classified with various current concept similarity computing methods in layer 0, and the metadata is classified with LMSMC algorithm in layer 1, and then the ontology mapping which combines several algorithms is made. Experiments indicate that the method performs better in improving precision and recall compared to individually using those algorithms.