W3C home > Mailing lists > Public > public-wai-ert@w3.org > October 2005

Re: Structure of Evidence class

From: Charles McCathieNevile <chaals@opera.com>
Date: Sat, 01 Oct 2005 20:02:31 +0200
To: shadi@w3.org, public-wai-ert@w3.org
Message-ID: <op.sxza2hq4wxe0ny@widsith.local>

On Wed, 28 Sep 2005 17:11:31 +0200, Shadi Abou-Zahra <shadi@w3.org> wrote:

> Right now the proposal for earl:evidence suggests the class to be a  
> collection of assertions. This will make queries (for example to compare  
> if two evidence clauses are equal) difficult because the order of the  
> assertions defines the actual RDF graph.

Right. You have to query whether things are in the list, and there is no  
direct support for that in SPARQL, so you have to get the list, and  
traverse it checking for the thing you want.

> Another approach is to define a property with range earl:evidence and  
> domain earl:assertion. This property will describe a relationship  
> between the evidence and the sub-assertions. Here is an example:

Minor correction - you have to make a class that has the properties, since  
evidence is meant to be a property of either assertion or result. The  
easiest way is to use a blank node, although it would make sense to define  
a class for evidence. So you would have

> <earl:evidence r:parseType="Resource">
>   <earl:hasAssertion rdf:resource="someAssertion"/>
>   <earl:hasAssertion rdf:resource="anotherAssertion"/>
>   ...
> </earl:evidence>

or you could have (incorporating the idea below of having both structures  
- we could select one or the other)

     <e:EvCollection r:ID="someID">
       <e:fullEvidenceList rdf:parseType="Collection">
          [each assertion]
       </e:fullEvidenceList" />
       <e:hasAssertion r:resource="someAssertion"/>
       <e:hasAssertion r:resource="otherAssertion"/>
       <e:hasAssertion r:resource="moreAssertion"/>

> What do you think of this approach?

The problem with it is that due to the open world nature of RDF, people  
can add a statement to your claim. A collection is, I am pretty sure, the  
only RDF construct where there is no way to extend it without throwing up  
an error.

One way to solve the problem is to require that there are both structures.  
This seems like bloating the format in favour of simpler processing of a  
query. It should not be technically difficult to implement. I think an  
evidence statement is an atomic thing - once you have made an inference  
its conclusions should not be altered. So if you change your mind you make  
a new assertion, rather than changing the old one.



Charles McCathieNevile                      chaals@opera.com
          hablo español - je parle français - jeg lærer norsk
         Web dreams are free:   http://www.opera.com/download
Received on Saturday, 1 October 2005 18:02:50 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:55:53 UTC