Expressing high-throughput data in a restructurable, integrated form for knowledge extraction

The growth of high-throughput and combinatorial methods in experimental materials science has pushed human-mediated data processing traditions beyond their limit. In order for such tools to be useful, automated data processing must become an integral part of the scientific workflow. Here we report on components of our scientific data management system, OpenMat, and especially a core component, AutoDB, that provide a foundation for creating scientific knowledge bases to enable data mining. In particular, our system implements the extract-transformload (ETL) paradigm in a flexible way that is specifically designed to organize heterogeneous scientific data for purposes of subsequent knowledge extraction.