SWAN Ontology and Bibliography module with reference to Bibo

[Note-- "ODL" below should be read as OWL-DL or OWL DL, see http://www.w3.org/TR/owl-guide/ ]
See wikipedia FRBR

in SWAN
two manifestations related to one expression (work)

When we import the data from PubMed, it gives a reference to a documentation.

can create an expression even if only one manifestation-- but you do not have to.

The manifestation has all the metadata, the expression is just saying these two things are the same thing

It's just a link. No blank node, in SWAN everything has an identifier [though instance could be a blank node?]

Wanted to use this instead of Same As, because semantically they are not the same-- and have different DOIs– Document Object Identifiers.

There's a service providing metadata for DOI (that I sent to Paolo so i have somewhere).

already doing inferencing, has a reasoner
Swan Ontology, a piece of it, is not LDL

You cannot do [RDFlists]
it is not LDL

but the principle is simple-- double-linked list, the next property has an inverse property; and they are transitive. Ask for each item that is followed by B-C
Give me each paper with author name Tim Clark as last author
you can write in rool to use lastItem directly
if you ask for Tim Clark neither first nor last, it gets more complicated-- tedious rules to say no previous or no next.

In Protege you can write logic rules that are run against the reasoner.
Not like in SPARQL, directly run against the reasoner.
SPARQL doesn't care about reasoning

So if I make a property transitive or not that changes how the reasoner works [but must look at other things in

In short: If your threshold is RDF, things can be simpler than done here. This is LDL, which requires more.

SWAN's LDL can be translated to RDF.

Does OWL2 help?

Paolo: OWL2 introduces new things that are farther away from RDF but the serialization it can be done in RDF.

Me: What does it mean from the output?
Paolo: Bibliographic records are the same, the only thing that has to be managed is the linking thing

Stephane: We are able to say we have a field for author, and we can have as many authors as we want.

Me: Can you give weight in RDF (you can in Drupal!)

Duplicate this into a new

Stéphane: Extend RDF CCK to support multiple ordered values. Generalizable beyond SWAN?

Paolo: Yes... [not sure specifically for other things] But there are lots of ordered things in SWAN! [Genes, claims]

Paolo: I don't want you to overstretch your effort. People could interpret it in the wrong way.

RDF is much more shareable.

Community is RDF based not SWAN based.

Stéphane: But yes, RDF based and LDL based.

Paolo: If you can export all of the ontology, and authors just in RDF.

Stéphane: Should you have an option to say something doesn't have an order?

Paolo: It has to on the screen

I am talking as a friendly IT person not for SWAN

exporting RDF
export OWL DL

[Replace all LDL above with OWL DL, I think]

I FOAF Document can be anything.

In [RDF], every URI has a type
but in Owl DL cannot give [two the same]

but in RDF you can

In Owl DL you have a person and it is disjoined from dog. But Document cannot be disjoined from anything, because people have been using "document" to refer to people -- same URI for homepage and for person, and that's freaking wrong. I don't know how to fix that. Had to put a restriction

me: there's no excuse for that [once I confirmed that if they just had myhomepage.com#me it would be a sufficiently distinct URI.]

Also using URI of web site for Publisher
so using workaround of making "index.html" the homepage for the publisher (to keep distinct URIs)

me: Drupal doesn't even have a working http://agaricdesign.com/index.html by default :-(

RDF people say one thing
OWL people say another
can't make them agree on the same point

me: Can't you just use publisher.com#self for the identity URI to leave the home page clear?

Paolo: I can do anything I want, but I cannot make everyone else do it.

Stéphane: If I don't do the first step in RDF first, it will fail in OWL.

It's easy to fix on the Drupal sites in the URIs we produce.

Paolo: You are making decisions

me: and that's where we're in a great position. We're about to be the 800 pound gorilla in this game.

(I still like #self, even if you might jump to a weird place on the page. Just pick a word that makes sense.)

A document is the same as an organization or a person -- so so wrong -- in RDF.

When we want machines to understand a little bit better, they won't be able to. FOAF is broken now.

In OWL, then, they are all the same.

That's in OWL LDL, not full

If you don't put [restriction] all the properties will be everywhere.

Saw this problem with FOAF and went to Alan Rutenberg. He said yes it's broken, fix it. So I created [OWL version]

These sources coming in from RDF [are broken].

Stéphane: There's an issue.

Paolo: There's more than one issue.

Stéphane: There's one issue that I want to talk about.

Paolo: Provenance right now is expressed through Dublin core-- DC:creator. But it does not distinguish between authors, curators, reviewers... Then there is the two levels of Dublin core. One in RDF everywhere. Then DCterms

But annotation in OWL is expressed through annotation properties-- which at first had no subproperties and no domain or range.

OWL DL

In OWL2, they recognized

Starting in OWL2, DCTerms: CREATOR ---subproperty---> DC:CREATOR will be legal

but in the meantime, using a separate semantic space that is [not inferenced properly -- I'm stretching the boundaries of my understanding]

You have the record. Title, volume, ISSN, etc.
Then you have DC:creator. Creator of what?

Stéphane: what we're really interested is the author...

Paolo: It's not clear, it's fuzzy what DC:creator is.

But you can add explicit author to the volume.

But author in a record is referring to another level of abstraction than DC:creator

This is my problem with provenance.

I cannot use DC:creator in the record because it's not-- it's the contribution, the record being created by someone.

FOAF:Person http://paolo ---DC:Creator---> FOAF:Person http://mama

person has home page and DC:creator of home page is the person (http://paolo)

S: Before SWAN there was no ontology for publications?

P: There was [bibtex] now converted.

When I mix that citation ontology with the real world it is [messed up]

If I have a record of a person, my mama is not the dc:creator of the record.

Back on the volume-- author, contributor

Problem is most people don't use DC:Creator for metadata, use it for real people

In the record you don't have a problem, but in mixed things you have an author.

Bibo

Do you really care who created the metadata?

Yes. If it is from someone i know/trust

conference proceedings

For the mapping it would be easier to go flat
but for the other things you want to do, it will not be flat
SWAN - you cannot flatten everything down. Authors and such.

Create a new URI

For the journals the URIs are the ISSNs. Always for the journals.

For other things

Namespace is URN ISSN and the number

Pull in a

You could have human cleaning up data-- still a problem with

How do you deal with importing a citation? One citation, a link to one journal and four authors. One node for the citation, one node for the journal, four nodes for the authors. Just importing this, how does PubMed deal with the relationships?

Authors have no object representation.

Stéphane:

What if we standardize on RDF? What if instead of writing a wrapper for each input, we write a mapper from the PubMed XML to RDF, and always import the RDF graph.

These pertain to the fields in my site which are already mapped by RDF CCK

<> a site:Citation ;

site:title "The SWAN" ;
author could aim to an external URI or just A1
A1, A2
:A1 a site:Person (again a local content type)
site:name "Ben"
:A2

take all the instances, ensure there is a content type for it

The URI is here but we could even have another field for the URI, an external URI

So the point is take, transform any type of non-standard or different standard or proprietary data and transform it to RDF
that's still mapping to the specific things on th esite, but it could be a little standardized

XML doesn't have URIs – if it did it would be XML-RDF

can parse the RDF

The other stuff out there

Simile-- Potluck, takes two URIs and you combine the data

Resolution