Addressing Internal Consistency with Multidimensional Conditional Functional Dependencies

Conditional functional dependencies(CFDs) have recently been introduced as a novel approach for capturing the external consistency of relational data by comparing tuples. They define bindings of semantically related values that originate from flat domains. In real world applications often domains with multidimensional metadata have to be addressed and data are not only stored in relational databases. Since CFDs are not applicable to measure the internal consistency of a single tuple, we present an implementation of CFDs in ontologies that encapsulates CFDs from a physical data storage and allows for the detection of violations of the internal consistency of a tuple. Our implementation of multidimensional CFDs (mdCFDs) encourages the creation of bindings between (root) nodes of tree-structures, which drastically reduces the number of bindings that have to be created. We enable the usage of mdCFDs for proactive error prevention and reactive detection of inconsistencies. The usage of multidimensional structures supports the suggestion of qualitatively good repair operations for invalid data. As CFDs are limited to the comparison of several tuples, our approach allows the detection of inconsistencies in a single tuple. The method is designed for application scenarios in domains with existing domain knowledge that can be referenced as a closed discourse world. We show how this approach is being motivated by an epidemiological cancer registry.