113. Statistical organisations wanting to share data internally or between themselves, need to share and agree on the meaning (semantics) of that data. This includes concepts and definitions, most importantly units and populations, variables and codes.

114. An important future expansion of CSDA, therefore, will be a semantic framework that allows a common way of modelling and sharing the semantics of data and metadata. Such a framework could be used at statistical organisation level as well as community-wise, and for example provide the ability to:

  • share data and (semantic) models within and between organisations
  • create coherent views of data and metadata assets
  • build metadata-driven systems

115. In regard to CSDA, some specific use cases are:

  • data integration, including conceptual and instance-level alignments
  • conceptual data access, which is the ability to access (and query) data at the semantic level independently of implementation details
  • quality checking

116. The features expected from this semantic modelling framework are:

  • in line with GSIM;
  • a well-recognised standard, facilitating the reuse of semantics already defined by other authorities;
  • machine-actionable, at least for the metadata representation;
  • supporting multi-linguistics, for human consumption;
  • comprehensive enough to cover the whole domain of interest;
  • covering microdata, macrodata and metadata (including structural, quality & provenance);
  • prescriptive enough to enforce share and reuse by default at a global community level.

117. In reference to these requirements and use cases, and awaiting the development of the framework itself, CSDA strongly recommends the adoption of the Ontology Web Language (OWL2, a W3C recommendation) and associated vocabularies. The re-use of existing OWL vocabularies such as SKOS, PROV-O, Annotation vocabulary, DCAT and its applications profiles is also strongly recommended.

118. In addition, a role of "keeper of global definitions", or ontology custodian, should be defined in the CSDA management framework, at the highest level of the statistical community.


  • No labels