Schema matching and embedded value mapping for databases with opaque column names and mixed continuous and discrete-valued data fields

摘要

Schema matching is a key enabler for addressing data access and knowledge acquisition in this new era of data deluge. Streams from big data and heterogeneous databases produce huge demands for analytics and information discovery. In that sense, this paper represents very important progress, since it provides an algorithm for schema matching based on two key aspects from different databases. Continuous attribute matching and the use of value mapping can point the way to enhanced schema mapping. The challenges have been addressed with a global objective function minimization algorithm that matches columns with continuous value attributes, modeled with a Gaussian mixture model and an iterative descent algorithm that embeds value mappings to enhance schema matching accuracy.