Re: Publication of RDF Data Access Use Cases and Requirements

The DAWG and the new document appear to point to possible solutions to a use
case I and others are working on. Our focus is on content rating for the
purposes of child protection but there are wider applications for the
ability to apply a single description to multiple resources.

The example data at http://www.icra.org/press/labellingWG/labelling2.rdf
includes "ratings" for various types of resource. One idea under discussion
is that a single document like this would carry all the rating data for a
domain, or group of domains. Resources on those domains would all point to
it but the RDF instance itself  would only be downloaded once and be
processed locally by a client such as a filter. Acceptance of requirement
3.5 (local queries) is good news!

Using the ICRA descriptive vocabulary, rating 1 would apply to one or more
chat areas on a website/suite of websites; rating 2 covers reviews of
action-adventure films; rating 3 is a rating for actual movie clips (this
includes a non-ICRA predicate) and at the end of it all there's a default
rating for everything else not specified above.

Which rating should be applied to which resource is defined by matching the
resource's URI against a pattern - either a simple string/substring-matching
pattern or a Regular Expression.

This makes Design Objective 4.2 important:

[4.2 Provenance
It should be possible for query results to include source or provenance
information.

Status: Pending.]

The provenance of the query would be an important piece of data not only for
matching purposes but also to prevent, or to explicitly allow, other
publishers to use the same set of ratings.

The optional match (3.6) discussion is interesting and might address an
obvious problem with this approach - that a URI can match more than one
pattern - indeed they all match the default. For filtering purposes we need
a single answer from the available ratings. Preservation of the order in
which the triples are returned would take care of this, i.e. the client
would always apply the first match. Additionally, perhaps the default rating
could be returned only if there were no match.

The use case is almost identical to the first and simplest of all, the
retrieval of an e-mail address, but would the desire to be able to match
against an object rather than a predicate constitute something new?

Our work in this is only just getting going so we're very open to comments!

Phil Archer
Internet Content Rating Association


----- Original Message ----- 
From: "Kendall Clark" <kendall@monkeyfist.com>
To: <www-rdf-interest@w3.org>
Cc: <eric@w3.org>
Sent: Tuesday, June 08, 2004 4:00 PM
Subject: Publication of RDF Data Access Use Cases and Requirements


>
> All,
>
> I'm happy to announce the first publication of the RDF Data Access
> Working Group:
>
> RDF Data Access Use Cases and Requirements
> http://www.w3.org/TR/rdf-dawg-uc/
>
> The Working Group is soliciting feedback from members of the Semantic
> Web and wider developer and user communities on this document and on
> the design it suggests.
>
> Please direct that feedback to <public-rdf-dawg-comments@w3.org>.
>
> Best,
> Kendall Clark

Received on Wednesday, 9 June 2004 08:04:12 UTC