will there be any linking to Wikidata?

[egonw]

For example, for the journals this would be nice, as many of them are in Wikidata already (thanks to projects like Wikicite)? To get an idea what is in there, check out @fnielsen's https://github.com/fnielsen/scholia ...

Source: https://tools.wmflabs.org/scholia/venue/Q180445

[lambdamusic]

Hi Egon, most definitely! We're just about to release some initial mappings, we'll keep you posted about that.

[egonw]

hi @lambdamusic, I was just talking to @SamWinthrop about this... we have some ideas in mind. The first step would be to get a Wikidata Property so that Wikidata can link to your graphs... a formatter URL (for webpage and for RDF) would be useful...

[lambdamusic]

hi @egonw just looking at this again. Don't know exactly what you mean by 'formatter URL' but maybe the redirect API could do? http://scigraph.springernature.com/explorer/api/#redirect
Essentially you can retrieve the SciGraph URI by passing the journal ISSN eg http://scigraph.springernature.com/api/redirect?issn=2365-631X

[pigsonthewing]

A formatter URL is used in Wikidata to specify the pattern for linking identifiers. for example, instead of storing the full URL of an individuals ORCID iD:

https://orcid.org/0000-0001-5882-6823

we store just the ID value (i.e. the URL slug):

0000-0001-5882-6823

and the formatter URL:

https://orcid.org/$1

By replacing the "$1" in the formatter URL with the stored ID value, we can build the full URL again.

This way, if a site changes its URL structure, we only have to make one change.; and for some sites we can store multiple formatter URLs (or indeed, formatter URLs on third-party sites).

[lambdamusic]

Got ya. Thanks for the explanation.

I suppose then that if you have a DOI (or ISSN, ISBN) in wikidata you can consider it the ID value, which then can be used with this formatter URL:

http://scigraph.springernature.com/api/redirect?doi=$1

Of course for an ISSN you want issn=$1 etc.. as per the docs .

This approach would be resilient also in case SciGraph URIs happen to change in the future.

[egonw]

@lambdamusic, I think what we need is formatters based on the IRIs used for journals and articles... along with the mappings of SciGraph<>ISBN, SciGraph<>ISSN, SciGraph<>DOI, of course... something like:

http://scigraph.springernature.com/things/journals/$1
http://scigraph.springernature.com/things/articles/$1

[lambdamusic]

@egonw I don't have enough context for judging whether one approach is more suited than the other. Surely though, what you are suggesting is more laborious than using redirect URIs, and probably more fragile too (for the reason I mentioned above).

Anyways, here's some more info:

• The complete list of journals IRIs can be found by downloading the 'journals' package on http://scigraph.springernature.com/explorer/downloads/.
• Worth mentioning that the SciGraph model defines two levels of abstraction for publications-type objects: the sg:Journal and sg:JournalBrand classes respectively model the product entity (similar to frbr:Expression or bibframe:Instance) and the more abstract work (similar to frbr:Work or bibframe:Work). So you need to figure out which one you want to link to based on Wikidata's model.
• As per the formatter, our instances URIs follow a regular form: http://scigraph.springernature.com/things/{{class name in plural form}}/{{scigraph ID}}. So based on the object you're mapping to, it should be pretty straightforward to determine whether it's articles or journals or journal-brands(note: multi word names have a dash). If you're not sure, just ping us or use the explorer to see some examples.
• Finally, SciGraph IDs are generated using an algorithm that takes into consideration a number of variables - some of them not publicly available. Hence your best bet is to get the URIs directly from the journals dataset mentioned above or via the common-identifiers redirect API.

Hope this helps!

[egonw]

Well, that is why I was wondering if the SciGraph IDs were stable and public... if they are not, then ISSN/DOI is what is left, but then we don't need a Wikidata predicate...

For the semantic web layer, do you have owl:sameAs relations between the Scigraph ID-based IRIs and an IRI based on the DOI or ISSN? Otherwise, there would not be a semantic link possible from the Wikidata side... and we'd rely on SciGraph to make links to Wikidata (which is perfectly fine, of course...)

[lambdamusic]

Publications have a property sg:doiLink pointing to crossref DOIs. Eg see http://scigraph.springernature.com/things/articles/e4fe974547d30d62aeabe83de034738d#table

So, not owl:sameAs, although it's trivial to generate it via a rule if that's what you need.

SciGraph URIs are meant to be stable and public. But other than a firm intent to accomplish that, I don't know what else to say!