Sorry to be “backseat driving” or “armchair reviewing” here, but I just saw the parent task and have a few comments on the markup proposed there. Feel free to split this into subtasks, ignore everything you don’t agree with, or mangle this task beyond recognition to better fit your workflow :)
- Since the JSON-LD contains the article link as the url and not as the @id of the JSON object, all the triples described in it actually have a blank node as the subject (see the N-Quads output in the JSON-LD playground). In theory, this means that all the other points I mention here don’t matter as much, since this describes a different subject than the sitelinks in Wikibase’ RDF export, but it’s still a bit odd.
- While the schema.org documentation claims that http://schema.org/ and https://schema.org/ are both fine, in general, RDF URIs must match exactly in order to describe the same resource (including the protocol), and most RDF-based tools will not recognize https://schema.org/Article as the http://schema.org/Article resource they’re familiar with. In Wikibase, we declare the schema: prefix to mean http://schema.org/.
- The JSON-LD connects the article and its Wikibase item via schema:sameAs, whereas in the Wikibase RDF export, they’re distinct entities, connected via schema:about.
- When the JSON-LD is converted to other RDF syntaxes (see the playground link above), the datePublished and dateModified values get the data type schema:Date, whereas in the Wikibase RDF export, the dateModified value (of an item – we don’t emit it for sitelinks) has the data type xsd:dateTime. But the schema:Date type doesn’t seem to be specified in the JSON-LD itself, so I’m not sure if this isn’t a bug (problem?) with schema.org itself, actually.
- In the Wikibase RDF output, the article title (schema:name) is a language-tagged string ("World"@en), whereas in the JSON-LD it seems to be a plain string ("World").
- The schema:headline is not part of the Wikibase RDF output, but should probably also be language-tagged in the JSON-LD. In fact, it might make sense to declare a language for the whole @context, though in that case the author and publisher names should be reset to English if they’re not translated (are they?).
In general, I’m approaching this from a very RDF-based perspective, whereas you seem to be more oriented towards search engines – for all I know, any of the points I raise here could actually hurt SEO, so don’t follow me blindly here :) it would probably also be good to get some input from other people more familiar with RDF and especially Wikibase’ RDF output, such as @Smalyshev.