Define mapping mismatch vocabulary

[nichtich]

For #194 a list of predefined mismatch types should be defined.

[nichtich]

The MDL vocabulary has been proposed by Reitz (2010) in A Mismatch Description Language for Conceptual Schema Mapping and Its Cartographic Representation https://doi.org/10.1007/978-3-642-15300-6_15 for data models / ontologies:

• Scope Mismatches (scope of source and target don't match)
  • Subsumption/Abstraction Mismatch
  • Overlapping Scope Mismatch
  • Categorization Mismatch
  • Aggregation-Level Mismatch
• Relation Mismatches
  • Structure Mismatch
  • Constraint Mismatch
  • Attribute Type and Encoding Mismatches

The vocabulary has since been extenden and is used in the HUMBOLDT Alignment Editor, see source code:

• Abstraction Mismatch
• Overlapping Scope Mismatch
• Incompatible Scope Mismatch
• Delineation Mismatch
• Categorization Mismatch
• Aggregation Level Mismatch
• Structure Mismatch
• Constraint Mismatch
• Attribute Assignment Mismatch
• Attribute Type Mismatch
• Schema Language Mismatch

For vocabulary mapping, the types of mismatches are likely different because we don't have relations but only map classes. In short we have three types at least:

• mismatch (general, this seems to be Scope Mismatch in MDL)
  • wrong target (this seems to be Incompatible Scope Mismatch in MDL)
  • wrong type

For "wrong type", a better type could be proposed:

• target is broader => mapping type should be "narrow match" instead or better map from a broader source concept instead (this seems to be one case of Aggregation Level Mismatch in MDL)
• target is narrower => mapping type should be "broad match" instead or better map to from a more specific source concept (this seems to be one case of Aggregation Level Mismatch in MDL)
• concepts only overlap partially => change mapping to to close match, related match or remove mapping (this seems to be Overlapping Scope Mismatch in MDL)

In summary, four mismatch types could cover the spectrum:

1. (scope) mismatch : mapping is wrong
2. wrong target / incompatible scope : better change target
3. level mismatch : mapping should use proper broader/narrower relation or better map on a different level
4. partial overlap : type should be close or related match

[stefandesu]

@UmaB7 also suggested two additional entries:

• Deprecated source concept (= there were changes to the source concept itself that make the mapping wrong)
• Deprecated target concept (= there were changes to the target concept itself that make the mapping wrong)

[nichtich]

To start with some URIs and labels:

• https://uri.gbv.de/terminology/mismatch/scope "scope of source and target concept don't match" / "Mapping passt nicht"

• https://uri.gbv.de/terminology/mismatch/level "wrong level of abstraction (too broad or too narrow)" / "Falsche Abstraktionsebene (zu allgemein oder zu spezifisch)"

• https://uri.gbv.de/terminology/mismatch/incompatible "incompatible scope of source and target concept" / "Quell- und Zielkonzept sind nicht vergleichbar"

• https://uri.gbv.de/terminology/mismatch/overlap "source and target concept overlap partially only" / "Quell- und Zielkonzept decken sich nur unvollständig"

[stefandesu]

I added a first version of the vocabulary to jskos-data: https://github.com/gbv/jskos-data/tree/master/mismatch

All further changes should be done there.