- From: Karen Coyle <kcoyle@kcoyle.net>
- Date: Sat, 1 May 2021 14:33:41 -0700
- To: Andrea Perego <ndr.prg@gmail.com>
- Cc: W3C DX WG - Public <public-dxwg-wg@w3.org>
Andrea, "Shape" is something that we've struggled with,[1] and I'll pass along your comments. It really is a node in RDF. The node often (but not always) represents an entity in the metadata, like "person" or "address", that gets described by properties. One way to think of it is that the shape is an organizer for a group of properties. In SHACL terms the TAP shape is most likely a target node. An example (a very simple one) would be useful here I think. Here's a shape with two properties: Bookshape dc:title dct:creator which could describe metadata that looks like: <http://example.com/book123456> dc:title "DC TAP" ; dct:creator "Andrea". Is some of the confusion from the fact that the shape is on the same row with the properties? In TAP, the shapeID is a value in a column for each of the properties for that shape: Bookshape | dc:title Bookshape | dct:creator which says that properties dc:title and dct:creator are part of the Bookshape shape. (Some prefer not repeating the shapeID on each row because it is easier to read, and will assume that programs processing the CSV will read the blank cell as "same as above", much like Open Refine or other tools.) There has also been a request that shapes be on their own row, which would allow you to express, for example, cardinality for the shape itself. So you would have: Bookshape | (mandatory=true) Bookshape | dc:title Bookshape | dct:creator or Bookshape | (mandatory=true) | dc:title | dct:creator However, we had trouble fitting that into a standard CSV without adding many more columns, e.g. cardinality columns for shapes could be functionally different from cardinality columns for properties. It is possible we were being too perfectionist, but that's how it worked out. Although not (yet) set in stone. I'm totally available to chat if anyone wants to. Some of this is hard to do in email. And I will pass along your comments to the group. kc [1] https://github.com/dcmi/dcap/issues/73 On 4/29/21 2:25 PM, Andrea Perego wrote: > Thanks a lot for the heads up, Karen! > > Reading the documents, there is a point not completely clear to me, > namely, what a "shape" is intended to be in DC TAP - in particular, > when applied to an RDF vocabulary. Is it a class (in the RDF sense)? a > constraint (in the SHACL sense)? both? > > I think one of the reasons for my confusion concerns the examples, > where shapes are denoted by literals (courses, tutors, etc.), whereas > properties by IRIs from vocabularies as Dublin Core, FOAF, etc. > > Andrea > > On Thu, Apr 29, 2021 at 8:13 PM Karen Coyle <kcoyle@kcoyle.net> wrote: >> >> All, >> >> I wanted to let you know that the Dublin Core Tabular Application >> Profiles specification is now a draft [1] on the DC site, and there is a >> blog post with additional information and links. [2] >> >> We think that this draft is fairly stable, but we are now working on >> various extensions including a manifest. The manifest document will not >> be a single solution but examples of various ways that additional >> information about the table can be expressed. >> >> What may be of interest to this group in that regard is that once we've >> gotten further on our ideas for a manifest we intend to analyze >> compatibility with the PROF vocabulary, and with the JSON context of CSV >> on the web.[3] >> >> I also intend to try encoding DCAT-AP in TAP, although it is dauntingly >> large so I may just do a portion. In any case, I'll get back to you when >> I have more to report. >> >> Thanks, >> >> kc >> >> [1] https://www.dublincore.org/groups/application_profiles_ig/dctap_primer/ >> >> [2] https://www.dublincore.org/blog/2021/draft_tap_specification/ >> >> [3] https://www.w3.org/TR/tabular-data-primer/#handling-language-in-csvs >> >> -- >> Karen Coyle >> kcoyle@kcoyle.net >> http://kcoyle.net >> >> > -- Karen Coyle kcoyle@kcoyle.net http://kcoyle.net
Received on Saturday, 1 May 2021 21:34:44 UTC