- From: Graham Klyne <GK@NineByNine.org>
- Date: Fri, 24 Jan 2003 18:02:35 +0000
- To: Libby Miller <Libby.Miller@bristol.ac.uk>
- Cc: "Seaborne, Andy" <Andy_Seaborne@hplb.hpl.hp.com>, "'public-esw@w3.org'" <public-esw@w3.org>
Libby, My take was that Andy's model/vocab catches this quite neatly, because what I read it to say way that the result set thus described can be asserted to be a correct result set [for a given knowledge base and query]. A subset of that, described in RDF by a subgraph of that given, is also a valid result set -- if it weren't, I'd have been concerned that the RDF semantics were being violated, in particular the subgraph lemma. With Andy's vocabulary, I think one could also have a concept of a maximal result-set, of whose graph all valid result-set graphs are subgraphs. Do you see where I'm going here? I'm just about to shut down for travelling -- I don't know if I'll be able to continue this exchange over the next week. #g -- At 03:42 PM 1/24/03 +0000, Libby Miller wrote: >I guess I mean that if you were associating a resultset with a query, >there might be several different resultsets that would be ok. by this I >mean several tables which are valid depending on whether the KB does >transitive closure on classes or not etc. Maybe it does this: I find it >difficult to read N3, and RDF schemas in general. Examples are the thing >for me, so I guess I should get on and try and add to yours from the >data we have. > >Many thanks for doing this Andy :) > >Libby > >On Fri, 24 Jan 2003, Seaborne, Andy wrote: > > > Libby, > > > > > you might also get multiple valid resultsets per query, > > > depending on the > > > power of the KB. > > > > I'm not sure which way round you mean "multiple valid resultsets per > query": > > the vocabulary allows multiple solutions per result table. And also > > multiple result sets per result graph because it is rooted from a single > > node. The example uses <> as that node but there is no reason it has to be > > that; you could have a bNode there, and have another starting bNode > > somewhere else. > > > > Could you give an example of when there would be multiple result sets? I > > can image a "query request" to actually consist of a series of > "queries" all > > of which should be executed. > > > > Andy > > > > > -----Original Message----- > > > From: Libby Miller [mailto:Libby.Miller@bristol.ac.uk] > > > Sent: 24 January 2003 15:05 > > > To: Seaborne, Andy > > > Cc: 'Graham Klyne'; 'public-esw@w3.org' > > > Subject: RE: Vocabulary for result sets > > > > > > > > > > > > you might also get multiple valid resultsets per query, > > > depending on the > > > power of the KB. > > > > > > On Fri, 24 Jan 2003, Seaborne, Andy wrote: > > > > > > > > > > > > -----Original Message----- > > > > > From: Graham Klyne [mailto:GK@NineByNine.org] > > > > > Sent: 23 January 2003 21:12 > > > > > To: Seaborne, Andy > > > > > Cc: 'public-esw@w3.org' > > > > > Subject: Re: Vocabulary for result sets > > > > > > > > > > > > > ... > > > > > > > > > > > > > > Hmmm... I wonder of there should be links to, or > > > identifiers of, the > > > > > knowledge-base and query used, so that valid results from > > > > > different queries > > > > > can be differentiated. In practice, I think this kind of > > > > > testing is a > > > > > relatively closed-world activity, so maybe it doesn't matter. > > > > > > > > > > > > > Graham, > > > > > > > > Good point. A number of properties to annotate the result > > > set would be > > > > good. Of course, nothing stops any properties being added > > > ... but putting > > > > them in the vocabulary encourages their use. > > > > > > > > Are there any suitable properties from other vocabularies to reuse? > > > > > > > > Also - this could be the result from a query, not just > > > recording information > > > > for a testcase. In this case, we still have a > > > query->single graph approach > > > > but the presentation of the results isn't a subgraph of the > > > original KB, but > > > > an encoding of the variable bindings. Each solution can be > > > substituted into > > > > the pattern for the query to generate a sequence of > > > subgraphs, each of which > > > > satisfy the query but the result set graph does not feel > > > like knowledege > > > > extraction anymore. > > > > > > > > Andy > > > > > > > > > > > > > > > > > > > ------------------- Graham Klyne <GK@NineByNine.org>
Received on Friday, 24 January 2003 12:54:23 UTC