Lost in datification? The journey of data from the primary source to the final interpretation

In the field of Digital Humanities, recent attention was given to the relationship between RDF triples and natural language in the context of natural language to RDF conversion of humanities texts. The rigid structure of ontologies obliges scholars to make critical choices during the formalization of data resulting from an interpretation of the cultural resource. This may result in crucial differences between the final RDF formalization and the natural language text in terms of how much of the final semantic content retains the original one and how much remains hidden due to the framework's formal structure. The verification of the adherence of structured data to the primary source is useful to test if the data model returns the semantic expressiveness of the primary source in order to pursue the specific goal driven by the computational process. In the context of humanities, in which the precision of information is fundamental for addressing sound analyses, this verification becomes crucial when derived data are treated as a substitute for the primary source in computational tasks. In this talk, we propose a three-step approach to verify the extent to which the RDF triples represent the respective content of the textual source from which they were generated within the limits of the modeling adopted. The approach is thus tested by proposing two case studies taken from two different cultural domains, namely literature and art history.