Added shapes for expressing literature references, based on schema.org

[apdavison]

Needed for expressing the provenance of parameters and data taken from the published literature

[MFSY]

I would recommend to put all schema.org related shapes in nexus-schema module for now.
We need to externalize the schema.org related schemas in BlueBrain/nexus-schemaaorg
I've opened an issue to do so (#46)

[apdavison]

The reason I put these shapes in a "literature" module is that I planned to add some related, but non-schema.org-based shapes, such as "LiteratureReview" (a prov:Activity), but I'm happy either way.

[MFSY]

In that case you can put schema.org related shapes in nexus-schemas module and reuse/extend them in the new module (in neurosciencegraph organization) you want to create.

Also note that there are two domains for schema.org related shapes:

• schemaorg: this domain reuses the shapes defined below and links them with a target declaration (e.g. targetClass: "schema:Book")
• schemaorgsh: this domain only contains shape definition without target declaration because of many reasons. The most important one is (in short) to avoid validation chaos.

The same reason leads us to have a neurosciencegraph/commons domain where only shapes (without target declarations) are defined.

I'm working on a documentation of this pattern we came up with.

Note that there is a SHACL version of the schema.org markups made by Topquadrant and available here. Worth checking the license and reuse them in BlueBrain/nexus-schemaorg project if possible.

[apdavison]

great! I was wondering what the reasons for that were.

[MFSY]

The doi identifying an article can be provided using schema.org/identifier property which should conform to the nexus/schemaorgsh/identifier/shapes/v0.1.0/identifiershape .
The doi can be provided like:
String

"schema:identifier":"http://doi.org/10.1093/acprof:oso/9780195159561.003.0008"

Or:
IRI

"schema:identifier":{"@id":"http://doi.org/10.1093/acprof:oso/9780195159561.003.0008"}

Or:
Property value

"schema:identifier":{
     "property":"http://doi.org",
     "value":"10.1093/acprof:oso/9780195159561.003.0008"
}

[MFSY]

The pubmed ID can be provided in the same way as the doi above using the nexus/schemaorgsh/identifier/shapes/v0.1.0/identifiershape.

Since multiple identifiers are needed (doi, pubmed id) it may be useful to create dedicated properties for them just like schema.org did. For example an "isbn" property is provided as an identifier sub-property. I've opened an issue for this.

[MFSY]

I've opened an issue to deal with the aim of these shapes: report on parameters extracted from literature and express their provenance (#51).

[pafonta]

I am a colleague of @MFSY at @BlueBrain. I am working on frameworks to extract facts and parameters from literature. Consequently, I have been made aware of this PR and the fact that I am welcome to comment it.

My comments:

1. Good work!

2. Is there a reason to allow a DOI for a chapter but not for a book?

3. Some document haven't a DOI nor a PMID to identify them but still need to be identified. A part of our discussion on this can be found on [References] [Annotations] Unable to attach a PDF for a reference without ID (DOI or PMID) BlueBrain/neurocurator#65.

4. It could be interesting to add a property for the language of the document. In our context it might be always 'en' but in others other languages can be used. This information is relevant for text mining tools or (faceted) search.

[apdavison]

Thanks!

I agree with your points, and I'll take them into account when I find time to revise this PR. Thanks to the pointers to the neurocurator discussion.