Optimization and comparison of schema matching solutions

摘要

As the number of data schemas grows, the need to find correct mapping from one to another is also becoming more stressing. Due to the representational heterogeneity of schemas, the solution is far from trivial. The ultimate solution is yet to come, but several promising algorithms have already been published. We have studied and implemented numerous schema matching approaches. It has turned out that their accuracy is strongly dependent on their configuration parameter settings and the specific input schemas. However in the current literature these factors are not taken into account by the performance analysis. Hence our goal is to set up a universal and correct benchmarking method. We have also developed methods that enable their correct parameter adjustment and permit the unbiased comparison of their performance. These techniques incorporate sophisticated mathematical formulas. Furthermore an indirect approach is also offered, which should also ease the correct parameter adjustment. Eventually we have developed some computational approaches that have been implemented and tested. We have conducted several experiments, which were performed on different kind of test schemas and validated our algorithms.