Re: Google Base and RSS and RDF

There was a piece by Adam Bosworth (of Google) in a recent ACM Queue magazine
about RSS and it's broader applications [1].  One thing that impressed me about
this article is that it suggested (to me) RSS (or Atom?) as a carrier syntax for
RDF(-like) information (i.e. something that could map the RDF abstract syntax),
which has the advantage of widespread support in existing browsers.  It seems to
me that this creates a "higher point of departure" for getting Semantic Web
machine readable information used and deployed in the wider web.

Maybe this is indicative of some aspects of thinking behind Google Base?

I'm currently involved with R&D for systems supporting scientific research,
mostly bioinformatics related, using Web and Semantic Web ideas.  In this work,
my own preference has for some time moved away from using RDF and triple stores
per se for storing the "semantic data", but rather to leverage existing data
formats in conjunction with SPARQL (cf. Chris Bizer, Richard Cyganiac, et al
work on D2RQ [2]).  Somewhat inspired by [1], I'm thinking about possibilities
of using RSS/ATOM for creating a merge of information from multiple sources --
starting from a conventional RSS/Atom summary of existing resources, and
extending to incorporate aspects of the content semantics.




Danny Ayers wrote:
> On 11/27/05, Xavier Noria <> wrote:
>>On Nov 27, 2005, at 9:58, Jeremy Wong 黃泓量 wrote:
>>>Google Base BETA [1] is available on the Internet. I just wonder
>>>whether it is a triple store. If it is so, it'll be the largest
>>>application of the semantic web.
>>In my opinion it would be an application of the semantic web if:
>>    * Data aggregation was SW-based, instead of, say,
>>      screen scrapping-based.
> Google Base does have RSS 1.0 input, which is an RDF vocabulary (their
> format is broken right now but they've suggested they're going to try
> and fix it). Their input allows user-defined properties, which takes
> this beyond RSS and closer to RDF proper. It's a little unconventional
> in that the new terms appear in a Google namespace, and the input
> syntax is constrained, but in essence it means they will be using RDF
> pretty much as designed.
>>or at least:
>>    * The built triple store, if there's one, was accessible
>>      using SW-standards rather than a search web form.
> Yes, some machine-friendly route to the data would make a world of difference.
> Implementation-wise, it seems like they've got something fairly RDF
> triplestore-like behind the scenes, although apparently lacking things
> like type inheritance etc that comes in RDFS. A month or two ago a
> presentation appeared on the Web from one of the Google people (can't
> find link, sorry) which described a database setup that was fairly
> loosely-structured - I seem to remember every item had a datestamp.
> Maybe that's what they're using. .
>>Otherwise formally they may be using a triple store, but in my view
>>there would be no added value as far as SW is concerned compared to a
>>regular server-side relational schema.
> Ok, so their system isn't altogether Semantic Web-friendly, in that it
> doen't (yet?) provide data out, but they've got it right as far as
> using Semantic Web technologies to add value to an existing system.
> Cheers,
> Danny.
> (somem,    of that pasted from:
> )
> --

Graham Klyne
For email:

Received on Monday, 28 November 2005 10:52:30 UTC