“Set-of-Strings” Framework Application to Imperfect Data Fusion

This paper is dedicated to the description of techniques which have been proposed for the extraction of perfect (complete and consistent) data from relational databases (RDBs) which themselves contain only imperfect data. The RDBs in question are those which have unknown or partly known values of attributes in elements of relations and/or elements which contradict one another in the sense of breaking functional dependencies associated with the particular RDB. Any extraction process must presume the existence of selections of consistent subsets of relations that are maximal in the sense that inclusion of any one additional element from the rest of the relation in question makes such a subset inconsistent. The “set-of-strings” framework, designed for Big Data modelling and handling, is applied to create a solution to the task of creating such subsets.