dc-provenance

[11:14] <@DGarijo> http://wiki.bib.uni-mannheim.de/dc-provenance/doku.php?id=opm_examples

[11:15] <masque> Daniel: Not much real life data out there

[11:16] <masque> … First example is from First Provenance Challenge: http://twiki.ipaw.info/bin/view/Challenge/

[11:17] <masque> … Karma, OPMToolbox, Tupelo have example graphs, but not real data.

[11:18] <masque> … OurSpaces and eBioFlow: the OPM model is appatently used, but the data is not openly available.

[11:19] <@DGarijo> http://openprovenance.org/examples/mash.pdf

[11:24] <masque> Michael: Looking at that graph, there are a lot of owl:sameAs relationships. This gets into hot water. Individuals on both sides of a process now have the same identity.

[11:25] <masque> Daniel: Yes, this seems to lead to an invalid OPM graph. A process cannot use and generate the same artifact.

[11:26] <masque> … I also tried to model the Europeana example, but it can't really be done (in a way).

[11:27] <masque> … OPM is about evolution of the artifact, not metametadata.

[11:30] <@DGarijo> http://openprovenance.org/model/opmo

[11:32] <masque> Michael: One of the crucial forms of metadata provenance is something like: #BnF says that #Leonardo is the creator of #Mona_Lisa. OPM has trouble expressing this?

[11:33] <masque> Daniel: Yes, because this does not involve some kind of process or change.

[11:36] <masque> … OPM is an abstract layered model. You use OPM as a top model to describe evolution; on the domain level you can use other standardized vocabularies for description of artifacts.

[11:38] <masque> … In our case of metadata provenance in Europeana, we really try to communicate state of description sets, not the evolution of these description sets.

[11:40] <masque> … Which prompts the question: Is evolution of metadata records really in scope of this group?

[11:41] <masque> Michael: I think aspects of it are: linking of version chains

[11:44] <masque> … and capturing original source vs. syndication source, etc.

[11:45] <masque> … Seems related to evolutionary aspects expressed by OPM, but somewhat different in our context.

[11:47] <masque> … We have a very limited view of what counts as a relevant process.

[11:47] <masque> Daniel: If you are only stating something like the source, then OPM is not necessary. If you want to say something about the “how” of the process, OPM is needed.

[11:50] <masque> Michael: I now feel we can completely ignore the ontogenetical aspect of metadata (provenance) for our purposes.

[11:51] <masque> Daniel: There seems to be a complementary place for it; Prov XG/WG is more concerned about evolution of information.

[11:54] <masque> Michael: Looking at the proposed metadata annotation set in our domain model draft, this piece of provenance still seems to be a piece of _descriptive_ metadata, even if we state source or process. An OPM graph is not descriptive metadata. This seems to be the big divide.

[11:55] <masque> … Probably something can be crosswalked over or enhanced to take part in the evolutionary approach of OPM if needed.

[11:56] <masque> Daniel: Even the new Google metatags (original source, syndication source) are “only” metadata; the result is a statement of the source, but not an assertion of the form “x has stated that y is the source”.

[11:57] <masque> Michael: They are still a description of the document, not description of the data. The reason for this seems to be that the provenance is baked into the metadata level, not the metadata annotation level. So statements about the metadata are hard to do in this case.

[12:02] <masque> … So far we haven't stated in that sharpness that the account of evolution/genesis of metadata might be out-of-scope of this group; we have looked at “provenance as process” and “provenance as descriptive information”, but not that distinctly. Is think this is a great outcome for today. Meeting adjourned.