Automating data integration in adaptive and Data-Intensive information systems

Abstract

Data acquisition is no longer a problem for organizations, as many efforts have been performed in automating data collection and storage, providing access to a wide amount of heterogeneous data sources that can be used to support the decision-making process. Nevertheless, those efforts were not extended to the context of data integration, as many data transformation and integration tasks such as entity and attribute matching remain highly manual. This is not suitable for complex and dynamic contexts where Information Systems must be adaptative enough to mitigate the difficulties derived from the frequent addition and removal of sources. This work proposes a method for the automatic inference of the appropriate data mapping of heterogeneous sources, supporting the data integration process by providing a semantic overview of the data sources, with quantitative measures of the confidence level. The proposed method includes both technical and domain knowledge and has been evaluated through the implementation of a prototype and its application in a particularly dynamic and complex domain where data integration remains an open problem, i.e., genomics.