2005-11-25

Tom Gruber on Tag Identity

Tom Gruber is famous for his striking and sensible aphorisms about ontologies such as "Every Ontology is a Treaty". He strikes again with this paper proposing an Ontology of Folksonomy.
In section 5, just before the conclusion, a bunch of questions about Tag Identity, showing that both Tags and URIs meet similar identification issues, in fact common to any naming mechanism, and neither is providing killing answers.

2005-11-21

Identity, Reference and the Web

A challenging workshop to be held in Edinburgh in May 2006. I've been invited today by the co-chair Harry Halpin to participate in the Program Committee. From the "Goal and Theme" section of the description.
URIs are the primary mechanism for reference and identity on the Web. To be useful, a URI must provide access to information which is sufficient to enable someone or something to uniquely identify a particular thing and the thing identified might vary between contexts. There is no doubt that as mechanisms for identifying web pages the URI has been wildly successful. Currently, URIs can also be used to identify namespaces, ontologies, and almost anything. However, important questions are the interpretation and use and meaning of URIs have been left unquestioned ...
Exactly indeed, what we are about here ... Interesting to see also Pat Hayes in the co-chairs list. I remember that quite a while ago in a private communication, Pat had stressed the fact that identity issues had been "sadly overlooked" so far by current Semantic Web technologies.

Compact URIs : The CURIE syntax

I've started using this compact syntax in a project using IPTC specifications. Beyond the practical aspects well presented in the W3C note, it strikes me that CURIEs are good candidate hubject identifiers. For example ISBN:0201749602 is a compact URI which can be expanded automatically by various systems in so many URIs providing different views of the book, such as ...

New SPEK release

Although previous release had not yield much feedback so far, I've kept hitting this nail of views, aspects and perspectives and eventually produced a new version of SPEK vocabulary. The new release is as simple as can be, although the RDF schema is quite weird at some points. It introduces a spek:Description which is not formally defined as subclass of rdf:Description, but which looks really like it. RDF editors are variously happy with it. SWOOP beta 2.3 seems to yield the best results among all I tried.
Hubject is no more an explicit class in this version, because I've figured out that about any resource could be used as a hub. I've kept "spoke" as the property linking a view to the resource it describes, but changed the direction : the spoke is directed inward to the (hubject resource), not outward from the hub.
As an example, I picked the "Air Pollution" hubject as defined by Wikipedia, and four different views : a term in a glossary, a descriptor in a thesaurus, a category in a taxonomy, and a class in an ontology.