Closed ppKrauss closed 6 years ago
Have you looked at the Tag Library indices? Mulberry did a lot of work back with version 3.0 to add this kind of info into them. Not exactly mapping to schema.org/XXX, but they do capture a good amount of semantic info.
Thanks @jats-laura (!), do you have some links or references about this Mulberry "semantic iniciative"?
I'm just talking about the indices in the Tag Library. The one in Archiving would be the most comprehensive: http://jats.nlm.nih.gov/archiving/tag-library/1.1d3/alpha-index/alpha-index.html
If you're looking for author, for instance, it's not a JATS tag, but the index will tell you what is: http://jats.nlm.nih.gov/archiving/tag-library/1.1d3/alpha-index/alpha-a.html
author
- see <attrib>
- use <contrib>
- see @contrib-type
This info is included throughout the index.
Hum... As you say, this is "not exactly mapping"... In my opinion it is a "good clue", but far from Web Semantic solution or Linked data initiative.
One way of mapping JATS elements/attributes to schema.org types/properties is via an XSL transformation to HTML, e.g. those found in jats-to-html.xsl and its subcomponents.
@hubgit Good (!), yes, I think this is the way... Can I collaborate there with this "mapping objective" in mind? A suggestion:
jats-to-html.xsl
with more itemprop
's (you add a lot of good ones, need only a little more). Some RFDa ( itemprop
) can be Schema.org, but some need other (LOV or Wikidata vocabularies).itemprop
decisions.enhance jats-to-html.xsl with more itemprop's
Suggestions for extra appropriate schema.org properties would be appreciated.
Here's an example of the current output (ignore the errors, mostly - it's just showing that "headline" and "image" were recently changed from "optional" to "required").
About "map" in a CSV at Github, the suggestion is to express XPath semantics, and discuss (looking for consensus) each decision.
JATS-XPath | scope | URL1 | WikidataID |
---|---|---|---|
//ref-list/ref |
CreativeWork | https://schema.org/citation | <Q1631107 |
//ref-list/ref[element-citation/@publication-type = 'journal'] |
ScholarlyArticle | http://schema.org/ScholarlyArticle | <Q191067 |
//ref-list/ref[element-citation/@publication-type = 'book'] |
Book | http://schema.org/Book | =Q571 |
//front//article-title |
CreativeWork | http://schema.org/name | =Q783521 |
//surname |
Person | http://schema.org/familyName | =Q101352 |
Example of discussion: in article-title
the name
property is vague, the semantic Q783521 is exact, and with SchemaOrg perhaps http://schema.org/headline is an alternative (URL2) for name
.
See the table at a collaborative sheet here.
PS: there are a similar problem with similar solution-by-sheet at SchemaOrg.
"Semantic mapping" is a old proposal, but never really used with NLM or JATS... A lot of semantic are expressed in the JATS documentation and can be extracted/confirmed with JATS corpora (PMC and other repositories with an infinitude of samples). I think that "semantic map" can be an enhancement for reuse-community and reuse-software.
The counting-statistics of LOV database show that there are some de facto standards like DCterms and FOAF... And the increased use at Web of semantic with de facto standards, like Schema.org, show that perhaps "is time to use semantic"... So, "map JATS elements to semantic (to RDF or other)" is a feasible project, and is a realistic demand to enhance reuse.
Example: the content of an
<aff>
tag is the "name of an institution or organization" (documentation), so this content is a http://schema.org/Organization and, as related<contrib>
contributor ("...with which a contributor is affiliated"), a semantic described by http://schema.org/memberContributor in general is a http://schema.org/Person but can be also a http://schema.org/Organization .
PS: many other semantic relations can be expressed using https://www.wikidata.org/wiki/Wikidata (example Organization is Q43229).