[LRUG] Linked Data & Ruby
Matthew Willson
matthew at playlouder.com
Tue Jun 30 04:17:52 PDT 2009
>> What is missing is the real use case of semantic/connected data for
>> current
>> trend of web applications.
>
>
> We publish all our /programmes and /music data as RDF.
>
> And we're researching how to publish using an RDF RESTful application
> service layer
>
> URIplay are using /programmes rdf to build their service
>
> http://uriplay.org/
>
> Maybe developers have yet to realise the potential?
If the RDF is genuinely being used by unstructured / semi-structured
spider-like clients then great; if it's just being used as a
structured data source by structured clients, though, then it seems
like there'd be nothing particularly "semantic webby" about the use of
RDF as opposed to some other structured media type for the data in
question.
To me it seems one of the main issues with the semantic web concept is
the issue of trust. It seems those who are genuinely doing broad,
useful inference as a result of semi-structured spidering of the web
(Google?) are using heavily statistical, 'fuzzy' techniques to asses
trust, reputation, weight things accordingly, infer facts, beat SEO
tricks etc - but these seem incompatible with the very black-and-
white, logical inference model that goes with RDF databases, that
demands total trust if you're going to base inferences on someone's
facts. (what if they lie, what if they introduce an inconsistency into
my database?). A lot of it has the smell of the slightly-naive poster-
boy logic-driven AI techniques from the 80s and 90s which were rapidly
fell out of favour as we headed into the 00s.
Not that there aren't answers to some of these issues, I know... but I
think a lot of us still need a lot of persuading about practical
advantages of the semantic web at this point. A start, for me, would
be to replace RDF triples (noun predicate noun) with a 'quadruple'
which includes the entity asserting the fact (entity asserts noun
predicate noun), using some kind of secure open standard for identity
with a signature...
-Matt
More information about the Chat
mailing list