RE: PROV for SKOS : Inventory of change activities for SKOS

Hi Jean!

You can see the revision history representation for Getty AAT here:
http://vladimiralexiev.github.io/aat/index.htm#Revision_History

I used rdf:type as classes from PROV-DC (subclasses of PROV-O)
and dc:type as literals representing an application-level change type (e.g. "term added" vs "parent added").
The model is quite simplified compared to full-fledged PROV, which is IMHO horribly complicated, see
http://vladimiralexiev.github.io/aat/index.htm#PROV

The purpose of this info is just to inform consumers of changes, it doesn't let them replicate the changes.
This is high-granularity and "imprecise" on purpose, so as not to accumulate historic objects.
Getty regenerates the complete dataset every 2 weeks, so such historic objects are considered garbage for this case.

You can at least get a good list of editorial actions, and together with dc:description figure out their parameters.
E.g. a "merged" action carries a note: "Dominant: <this.term> (<this.id>), Recessive: <rec.term> (<rec.id>)".
It is also reflected in converting the Recessive concept (Subject) to an Obsolete Subject:
http://vladimiralexiev.github.io/aat/index.htm#Obsolete_Subject

Please note that this list reflects only actions of public interest. E.g. candidate (propose), approve, reject are not exported.

The granularity, object lifecycle and historic object accumulation are the crucial considerations for such a model.
Consider that a concept can be merged, umerged, deleted, undeleted, etc in any number of ways.
The model needs to define a comprehensive semantics of sequences of such actions, and perhaps of undoing actions. 
Can one undo actions in the middle of a sequence?

Cheers! Vladimir

Received on Tuesday, 3 June 2014 15:26:24 UTC