[Application-profiles-ig] schema profile for google recipes - Correction

Wed May 27 20:56:54 BST 2020

Correction, last paragraph: John Samuel, not John Roberts. (Man's name, added an "s" - I was close.)https://github.com/johnsamuelwrites/ShExStatementskc On 5/27/20 12:21 PM, Karen Coyle wrote: >>>On 5/27/20 3:21 AM, Thomas Baker wrote:>>On 2020-05-26 08:17, Karen Coyle wrote:>>>[3] This brings up another question that I will add the list of running>>>questions, which is: by default, is an AP open (in the RDF sense) or>>>closed?>>>While ShEx has a default of open, I think that the main purpose of an>>>AP is>>>predictability, and therefore "closed" seems to be the right answer.>>>>In the DC context, historically, application profiles started off>>largely as>>expressions of rough consensus within a community -- guidelines in the>>form of specifications published as Web pages rather than as>>validation schemas, as things that could easily be extended with extra>>properties rather than as rules saying "use these properties with>>these constraints and no others".>>>>In a ShEx schema (a collection of shape expressions), any particular>>shape can be marked as closed, which means: "Data conforms to the>>shape if it consists _entirely_ of triples using _these_ properties>>with _these_ constraints". By default, using a property in a ShEx>>shape "closes" that property in the context of the shape, so the>>property-value pair "rdf:type foaf:Person" closes "rdf:type" by saying>>it must be used only with the value foaf:Person -- unless the keyword>>"EXTRA" is added, in which case a person could be described with>>additional (and unspecified) outgoing rdf:type arcs such as "rdf:type>>ex:Human".>>>>>The ShEx defaults seem like a reasonable starting point: shapes open>>w.r.t properties, individual properties within a shape closed w.r.t>>value constraints. We could decide to articulate different default>>assumptions about openness for the purposes of the template. We might>>even be able to find a way to squeeze CLOSED and EXTRA into the>>template if we really wanted.>>>>However, I do not think it would be productive to try to engineer more>>subtle expressivity for what should be a very simple tabular format.>>I'd rather see us focus on completing a simple template, with>>examples, with well-articulated assumptions, and encourage users who>>need more to use the template as an on-ramp for schemas in more>>expressive languages such as ShEx. It is hard enough to explain these>>subtleties to ShEx users and would be no easier to explain to users of>>the template.>>>>Exactly. I would dislike baking this ShEx formalism into our simple>template. The use of "EXTRA" seems quite awkward to me, and I don't see>a logical way to get that information into our simple template in a>general way, much less to explain it to the creators of simple profiles.>In fact, I'm within an eyeblink of insisting that we forget ShEx, et al,>for the moment and develop our template to serve general purpose AP>development. It would be a huge mistake, IMO, to bake anything>ShEx-specific into our work. I would bet dollars to donuts that few if>any of the folks create Wikidata schemas could explain EXTRA nor why it>is needed. (Few of the schemas use it, yet more of them may actually>need it.)(*)>>Of the folks that I am aware of that are creating RDF profiles, none are>currently using ShEx, mainly because there is no ShEx application and>interface that they can make use of. Instead, some are using SHACL>because that is provided in TopBraid but most are creating their own>home-grown validation, which probably works best for simple metadata. My>preference is that we not design for any specific validation standards.>Among other things, they may not be suitable to our level of simplicity.>>If there does develop a community that wishes to define ShEx-specific>profiles, hopefully as an extension of our simple template, that would>be great. Ideally that would happen within the ShEx community group,>perhaps in coordination with, but not emanating from, DCMI.>>Meanwhile, we still have the open/closed general issue with RDF, so I'll>add that to our list of open questions. It occurs to me that it may fit>logically onto our "entity" row, because it will mainly be used to>determine if other properties are or are not allowed.>>kc>(*)顺便说一句,我开始看WikiData templates used by John>Roberts and they are quite different from the simple template we are>developing. I'm trying to create a template with the WikiData schema>data, and I can say that it is ending up being considerably different to>他设定的csv。这是不同的ence between>addressing a specific application and creating a general model. I'll try>to present it by the end of the week.-- Karen Coylekcoyle at kcoyle.nethttp://kcoyle.netskype: kcoylenet