User communities are expected to select a subset of the classes in the DDI-CDI model and to document the specific use of those classes for describing the data and metadata resources of interest to that community. Those implementation guides would include also the controlled vocabularies (including ontologies and/or classifications) to be used, as well as the syntax representations. Combined use with other standards would also be indicated. Some work has already been done in this area. The two examples identified by the WG are (1) the use of DDI-CDI at the UKDA for describing long, wide, and multi-dimensional data, and (2) the description of integration processes for climate and social data at Sikt. These examples would be published at the DDI Alliance site to show other communities how to produce their own implementation guides.
Currently, DDI-CDI has two syntax representations, in XML and in RDF. It is expected that the model will also be implemented in a variety of other ways. The basic approach to syntax mapping is to take the UML Class Model Interoperable Subset (UCMIS) profile of UML features and map those against the feature set of the target syntax as necessary. High-priority target syntaxes include Python, GraphQL, and SHACL/ShEx, although other languages such as R might also be considered.
The use of DDI-CDI in combination with other standards is expected, and the various ways in which this can be done require documentation. The Cross-Domain Interoperability Framework is one example of how DDI-CDI will form part of a combined set of standards, but other combinations will also be used. Significant standards include DDI Codebook and DDI Lifecycle, SKOS/XKOS, DCAT, Schema.org, PROV-O, RO Crates, and others. This group will look at how best to document these alignments, whether this means mappings in a format like SSSOM, relationships included in the UML model, or other forms of documentation.
DDI-CDI is designed to fill a gap in the metadata covered by other domain-independent standards, but the landscape of metadata standards, ontologies, etc. is a complex one. This group would look at positioning DDI-CDI in relation to other, similar models so that practitioners could better understand its purpose, and how best to employ it. This work would address not only the various types and levels of ontologies and metadata models, but would look at both domain-specific and domain-independent ones. The output of the group would be a white paper intended for metadata experts and implementers.
Standards such as I-ADOPT and the OGC Observations & Measurements describe the relationships which provide critical information about the reuse of measurements in terms of how variables are related. In some cases, there are dependencies between variables which are critical in order for them to be meaningful. Documenting such relationships and dependencies is important for cross-domain scenarios, and there is some support provided in the existing DDI-CDI model. This group will look at whether existing support is sufficient and optimal for expressing these relationships between variables, and whether DDI-CDI can serve as a vehicle for expressing standards such as I-ADOPT in a useful fashion.
There has been some initial discussion within the WG regarding the combination of traditional numeric and coded datums, and other non-traditional ones such as segments of free text, images, video, sound files, etc. This group would look at the utility and further description of such non-traditional datums within traditional data sets, and how the DDI-CDI model could be adapted to describe them. This is not an approach to describing qualitative data, but those data sets which combine traditional datums and those of a more qualitative character.
Dependencies between Topics
Implementation Guides have a connection to syntax representation (representing subsets in dedicated syntax representations)
Variable relationships may be impacted by non-numeric, non-coded datums
Arofan’s presentation from the morning of Mon 25 October can be viewed here.