About the model: Question 2

2. We need to be able to identify a work for human beings (rather than machines) using a combination of both the author and the title (when there is an author), but we also need to treat the author as an entity (in RDF terms, a class?) so that we can create a record for it that contains all of its variant names, biographical information and so forth. I think what I am saying is that we need to treat an author as both an entity in its own right and as a property of a work, and in many cases the latter is the more important function for user service. Is it possible to model this? Or is it possible that RDF (and other data modelling) works against effective use of bibliographic data because of an absolute requirement that something either be a class or a property, but never both?

OWL Web Ontology Language Guide (http://www.w3.org/TR/owl-guide/) p. 7 (Section 1.1 under OWL DL):  “A class can not also be an individual or property.” “A topic could be class and instance at the same time”–The Topic Maps Handbook by H. Holger Rath, on the web at www.empolis.com

“Sure. I guess this part points to something that splits the RDF and Topic Maps worlds apart. With RDF, you have an ever-growing tree-structure with no cross-pollination without what’s known as anonymous nodes. So, every thing expressed, every relationship, is created as a child-node of the thing it belongs to. Example ; 

Alex     is_a : person    

has_a : dog       

with_name : Oscar       

of_type : english_cocker-spaniel    

married_to : julie

Let’s take just the one part of this out, and express it as a falling triplet ;  

“Alex” : http://shelter.nu/me.html     

 “married_to” : http://some.ontology/human_relationships#marriage        

“Julie” : http://shelter.nu/my_family.html#julie

For each thing that’s attached to me I must make a triplet, and for every thing within a triplet more triplets to explain them. Now, the thing here is ; Where do you express these things? Where do you express that Julie is a person, and where do you define that URL?With Topic Maps, all things must first be a topic, so the same tree would be (and remember the Public Subject Identifiers talked about earlier, PSI’s) ; 

topic : “Alex”, PSI=http://shelter.nu/me.html 

topic : “married_to”, PSI=http://some.ontology/human_relationships#marriage 

topic : “Julie”, PSI=http://shelter.nu/my_family.html#julie

These things are expressed as things. Making the links between things are done *outside* their context, treating relationships as topics themselves ; 

association of type “married_to” : between “Alex” and “Julie”  (‘association’ is simply a topic of type ‘association’ 🙂

This association itself can have a URL (unlike RDF relationships), a PSI, a name, or more relationships and properties attached (basically, RDF’s ability to reify statements is handled ambiguously). The reason I’m talking about these RDF and Topic Maps differences is to point out this problem of where to define a thing. Where do you define a thing in RDF? There is the notion in the RDF world that you have a separate document with RDF triplets to define these things which you link in when you resolve or infer over your triplets, but in RDF this is implicit while in Topic Maps they are explicit. this is one of those things that’s confusing about RDF.

But back to your question ; Just talk about things, there and then. If they resolve to the right thing, your inference engine will thank you and be able to work with it. The persistence of these things lie with the URL you choose for your thing. If it’s an author, I can do ; 

“Frank Herbert” : http://authors.psi.org/f_herbert(1920-)

And do ;   “Dune”      is_a book : http://things.psi.org/book     

has_author : http://authors.psi.org/f_herbert(1920-)

It should be up to smart SemWeb systems to pick these URLs up and use them for persistence. Smart systems should also be able to link http://authors.psi.org/f_herbert(1920-) and http://authors.psi.org/f_herbert together as the same author, but this unambiguity lies at the heart of the semWeb problem.

RDF comes in many parts, and as basic RDF itself there is no such constraint (everything can be anything). You can build those constraints with RDFS. You can use more specific language within RDF to define things, or more complex statements with using the three levels of OWL. (At this point we’ve included five levels of ontology redirection, and I’m sure I’ve lost you along the way … :)”–September 11, 2007 email from Alexander Johannesen.

2 Responses to “About the model: Question 2”

  1. Martha M. Yee Says:

    Rob Styles sent out a very useful posting on this subject to the NGC4LIB list and gave me permission to copy it here:

    Martha, to answer the question if it’s still posed… A resource,
    such as http://example.com/composer/monteverdi in your examples may
    have a class (rdf:Type) and be used as a property by other resources.
    The distinction between classes and properties is in the schema URIs
    rather than the URIs in the dataset, so your example might look like:

    Monteverdi” .

    so the Class http://someschema.com/contributortypes#Composer is used
    to indicate that Monteverdi is a Composer, and the property http://
    purl.org/dc/elements/1.1/creator used to indicate that he composed
    whatever 123456789 refers to.


  2. marthamyee Says:

    It still appears to me (based also on a conversation with Bradley Allen) that we can model “preferred name” in an RDF model for all entities except work (and expression). For works and expressions that need to be identified using the name of the principal author in conjunction with the title, this preferred name convention cannot be specified in the model but must be left to the individual programs that are designed to work with the data (the same situation we have now in which systems people cannot seem to understand that the work/expression identifier (preferred name) consists of *both* the name of the principal author in the 1XX field in MARC *and* the uniform title of the work in the 24X field in MARC). If I am right about this, I think it could be argued that the RDF model is not yet sophisticated enough to handle bibliographic data.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: