Incremental Schema Mapping

Schema mapping that provides a unified view to the users is essential to manage schema heterogeneity among different sources. Schema mapping can be conducted by machine learning or by knowledge engineering approach. Machine learning approach needs training data set for building models, but usually it is very difficult to obtain training datasets for large datasets. In addition, it is very difficult to change the model by human knowledge. Knowledge engineering approach encodes human knowledge directly, such that the knowledge base can be constructed with limited data, but it needs time consuming knowledge acquisition. This research proposes an incremental schema mapping method that employs Ripple-Down Rules (RDR) with the censored production rules (CPR). Our experimental results show that RDR approach shows comparable performance with the machine learning approaches and RDR knowledge base can be expanded incrementally as the cases classified increase.