Re: ontology entity-instance backlinks

Just a note to add into this interesting discussion.

POWDER [1] is working to make some of this possible and OAI's ORE [2] 
works on similar lines. Both offer a way of adding triples to other 
types of content, notably Web pages. POWDER is essentially an XML 
dialect that makes it easy to add triples to loads of Web pages at once 
by associating predicates and objects with URIs that match string or reg 
ex patterns (everything on example.com or whatever) - and you can turn 
the whole thing into OWL by applying GRDDL (you need a semantic 
extension for OWL compatibility but it's pretty minor and we're close to 
releasing the Jena extension that does this).

Linkage to the POWDER documents is provided by any of RDFa mark up, HTML 
Link elements, or their HTTP Link header equivalent [3] (so we're not 
tied to HTML). Importantly, all POWDER documents are attributed to 
identified individuals (Agents) which is critical for decisions about 
whether to trust the data or not.

The problem, as has been identified, is that adding metadata/mark up to 
pages is time consuming and therefore expensive - even if you have good 
tools to do it. POWDER, at least, tries to address that by providing an 
efficient way to assign, say, licence data to loads of resources at once.

We're at TPAC this week and expect to release a second (and final) Last 
Call within a very short time with at least Proposed Rec in December. 
Give us a few more says to dot the Is and cross the Ts...

Phil.


[1] http://www.w3.org/2007/powder/
[2] http://www.openarchives.org/ore/1.0/primer
[3] http://tools.ietf.org/html/draft-nottingham-http-link-header-02


-- 

Please note my new e-mail address. My ICRA/FOSI e-mail addresses will 
not function after the end of November.

Phil Archer
e. phil@philarcher.org
w. http://philarcher.org/


John Flynn wrote:
> Yes, XHTML2.
> 
> I would propose that each participating web site include a top-level tag
> indicating it contained semantic markup. A dedicated web crawler would find
> those sites and build a search table containing the specific instance data,
> the referenced ontology for each instance, and the original web site uri to
> allow manual checkback to verify the data if desired. A semantic search
> engine would contain a list of applicable ontologies for a hypothetical
> search for some specific automobile information. The usefulness of those
> ontologies might be determined by some algorithmic which included frequency
> of reference, much like Google does. It might find instance data directly
> from the ontology and its own associated instance data and/or from finding
> the applicable ontologies in the search table and extracting the relevant
> instance data.  If my semantic search of information about automobiles
> returned specific data and I could see it came from www.bmw.com, I would
> probably accept that data on face value. If the data source was
> www.futurecars.com, I might want to visit the source web site to look at the
> context of the data. 
> 
> John
> 
> -----Original Message-----
> From: semantic-web-request@w3.org [mailto:semantic-web-request@w3.org] On
> Behalf Of carmen r
> Sent: Monday, October 20, 2008 1:47 PM
> To: semantic-web@w3.org
> Subject: Re: ontology entity-instance backlinks
> 
> 
>> developers to mark up their data in a way that it can simply be accessed
> via
>> the Semantic Web. Both RDFa and HTML2 are addressing this issue, but there
> 
> i assume you mean XHTML2? HTML5 also has some metadata stuff
> 
>> is still no simple way to html tag specific local web site data as
> instances
>> of a widely used ontology located at a remote site. You  might envision a
>> generally accepted ontology on a domain such as  "wine" that many of the
>> individual html web sites on that subject would link their data to as
>> instances. A capability to search that ontology could lead back to the
>> marked up instance data,
> 
> this is an interesting idea. are you proposing some sort of pingback service
> for ontology hosts?
> 
> 
> 
>

Received on Monday, 20 October 2008 20:50:25 UTC