All classifications included in Caliper have been converted into RDF, with the idea of reusing, anywhere possible, vocabularies and best practices already available for RDF. In short, given that RDF is a data model made of triples like "subject - predicate - object", we aimed reusing standard vocabularies for the "predicates", and adopting known best practices for all other aspects of the publication of data in Caliper.
Given the general principle of "sticking to the standard" as much as possible, we defined the models for the statistical classifications following an iterative approach. After a careful study of the classifications, usually in collaboration with domain experts, we defined a first version of the conversion, for the purpose of gathering feedback from the experts. Then, a cycle of interviews and fixes followed, until we reached a final product that was sound from the point of view of the domain expert, and also compatible with the requirements imposed by the tools adopted.
The actual conversion was made either with ad-hoc routines written in Python, or by means of the Sheet2RDF tool included in VocBench, sometimes a combination of both. The input format was generally a (set of) spreadsheets, occasionally database dumps.
Two standard vocabularies for RDF are at the core of the RDF versions of statistical classifications in Caliper:
- SKOS constructs are used to express the building blocks of all classifications, including their hierarchical structures. For example, a classification item is a skos:Concept, a code is a skos:notation, which the hierarchical relation between general-specific items is given by the properties skos:broader and skos:narrower. SKOS is a W3C Recommendation from 2009.
- XKOS is used to express cotnstuctus that are specific to statistical classifications, such as the notion of level, or of correspondences between classifications (rendered in a more fine grained way compared to the notion of match available in SKOS). XKOS is an RDF vocabulary endorsed by DDI.
Other standards used in Caliper include: