Experimental Discussion on Extracting Data Morphemes Based on Contexts and Variables from Heterogeneous Metadata

Abstract

In recent years, value creation through data exchange and distribution among different organizations has been attracting attention as a new source of innovation. However, the design and acquisition of data is highly specialized. Therefore, data exchange between others with different background knowledge involves risks such as the application of incorrect data analysis methods and inconsistencies during data integration. In this study, we believe that sharing not only the actual data but also the intrinsic meaning of the data is effective in addressing these issues. In this paper, we define "data morphemes" as the basic structure of data, and obtain the semantic units of data as the contexts and variables, and their relationships from textual information contained in metadata. We then discuss the results of our attempt to extract data morphemes using association analysis with network approach.