Talk:Q252485

p276

@Shonagon, El Caro:. This item is about the print in general, not about a particular copy, I do not think it makes sense to use location (P276) here. If we want to document individual British Museum or Met version, we can create about them, with p276, accession number and all. But documenting details about indidividual versions in a general item is rather confusing. I see that the French and Spanish Wikipedia articles provide location data in the infobox, but it seems pretty weird and arbitrary (why the Met version ?). --Zolo (talk) 08:52, 20 September 2014 (UTC)

yep Zolo, always the same problem of an artwork with his expressions, still without solution. We should separate each expression in specfic items (for location (P276), image (P18), inventory number (P217), depicts (P180), exhibition history (P608), ...). But how linking those items ? Using the subclass of (P279) would generate a huge mess and I think we really need a new property to solve this issue ; something like "version of", similar to edition or translation of (P629) for books. It could be a simple solution for everybody (contributors, skilled or not, and reusers) and for existing items and new items. Other similar cases : http://www.zone47.com/crotos/?l=fr&p=&nb=20&s=V%25C3%25A9nus+Adonis+titien Napoleon Crossing the Alps (Q59201), Aphrodite of Knidos (Q618535) Shonagon (talk) 09:41, 20 September 2014 (UTC)

@Shonagon: yes, this is a recurrent issue. It sounds a bit more like the edition -> individual book than the work -> edition relation (one item defines what the object looks like, and the other is about a material object). The parent item has to be about an idea rather than about a material object, so the logical relation between the two items seems to be "materialization". That seems pretty much the same thing as "instance of". So maybe it is fine to call The Great Wave off Kanagawa (Q252485) a subclass of print (Q11060274) and "Great Wave (Met copy)" an instance of The Great Wave off Kanagawa (Q252485) (that does not necessarily prevent using other p31 at the same time) ? I think it would be manageable this way, if we also tag The Great Wave off Kanagawa (Q252485) as an instance of "print template" and "Venus and Adonis by Titian" as a "painting concept" or something like this (those are ugly names but we would still need them if we created a new property). --Zolo (talk) 10:43, 20 September 2014 (UTC)

I've just asked a question about art infoboxes on French wikipedia (fr:Discussion_modèle:Infobox_Art#Plusieurs_localisations) and was about to ask a related one here on wikidata. So, thanks Zolo for pointing the issue here. I think your solution may be relevant. --El Caro (talk) 16:34, 20 September 2014 (UTC)

Thanks, I replied there. Note however, than things can be done a bit differently in Wikipedia and Wikidata. So something like Kassel Apollo (Q1607707) ? -Zolo (talk) 15:50, 21 September 2014 (UTC)

@Shonagon: Never forget that Wikidata is made first to be requested. With the subclass of (P279) solution for this issue, if you want to list the artowrks (and not the expressions) of a painter you will need to mix subclasses and instance in your query, it will be a huge mess. Even more because most of contributors and bots contribution will have difficulty to identify and to edit correctly those data, and of course to understand this data modeling. One other problem is as the wikidata is growing the arborescence of subclass of (P279) is more and more complex ; for example as a consequence I have stopped to request subclass of (P279) of drawing (Q93184), items are too much heterogeneous. Maybe I'm wrong but imho this subclass of (P279) solution will be a worst problem, both for contribution and reuse. As completary information, an article in French about the similar issue in frbr : http://rda.abes.fr/2013/08/30/frbr-de-lexpression-bordel/ In life a materialization is implicitly a conceptual artwork (a painting is a "painting concept" concept too) ; if it is necessary to make disctinction, as the samples above, instead of changing the modelisation, I would prefer creating this artorwk with the same instance of (P31) and adding a property which can be used to solve easily this issue for contribution and querying. Shonagon (talk) 21:31, 21 September 2014 (UTC)

@Shonagon:. I did not mean to suggest that we should not add "creator: Hokusai" on an item about a print of the Great Wave or instance of statue on Kassel Apollo (Q1607707) (actually I just added the latter). Indeed, for the time being at least, it seems better to add it. What I meant is that we should really differentiate items that are about a single material object and those that are not. For example, it makes maintenance more straightforward as we can put differnt constraints on the structure of items that are material objects and those that are not. I think P31 seems to be fit for it, but we can add several p31 values per item if needed.

For queries it seeems to work fine. Imagine the Met and the British Museum give their whole online database to Wikidata (!). Then we have at least to Great Waves expressions (that should ideally have data about their accession number, provenance, state of conservation, etc.) If we want to get prints by Hokusai depicting a wave, and if the item about the Great Wave in general has the same p31 as items about individual copies and we want to get a list of prints by Hokusai depicting a wave. We do not want to get a list of all copies of the great wave, and this is much easier to avoid if we have differnt p31 values. Other queries like "works by Hokusai in the Met" would still work the same way. And queries like "all works depicting waves" would still be messy but that seems hard to avoid.

All paintings may also be paintings concepts but given that our issues here are essentially practical, I guess we can avoid that. For prints, given the plate exists before the copy, things may be a bit different, but I am not sure this needs have an impact on Wikidata structure. --Zolo (talk) 08:20, 22 September 2014 (UTC)