Re: Use Case: DC-01

I've written a few metadata-based best match search engines
in the past, and so I've got at least a general clue as to
the challenges of a "best match" algorithm.

While I remain very interested in this use case, and agree
that it would be great if the DAWG solution would accomodate
it, I also see the great potential for months spent debating
just what is a "best match" with regards to some query aimed
at an RDF graph.

I have my own views, formed out of my own experience, but I
have no idea how closely that will correspond to others views.
Who knows, maybe there's a magical concensus hiding about here,
but I'd be very cautious about taking on this particular use
case, especially this early in the process, before the real
core is locked down.

I.e., let's perhaps add this to the issues list as something
to return to, and maybe even attempt to tackle.

Eh?

Patrick


On Apr 02, 2004, at 15:21, ext Dirk Colaert wrote:

>
>
>> This sounds like "best match", which would require some forumula
>> for ranking alternative results, which may be alot of work.
>>
>> I personally like this use case, and think it definitely reflects
>> real-world needs. I'm just wondering if it's (a) too big to take
>> on and (b) something that needs more research/exploration before
>> standardizing.
>
> I certainly agree that solving the query is a difficult task. But I 
> guess
> this is outside the scope of the workgroup. The point is do we need a 
> query
> language capable of:
> 1- giving constraints (no doubt I guess)
> 2- imposing a sort order on some criteria
> The difficulty is that that 'criteria' is not necessarily a data 
> element in
> the rdf document but implied in the rules within the document (as you 
> said:
> "best match").
>
> The relevance of the Use Case is that it points clearly to that 
> capability.
> It is up to the WG to decide whether or not we want the QL to have that
> functionality.
>
> My point is: The problem is very common so, if we can't express this 
> problem
> with SQL, if we can't express the problem with xpath/opath (object 
> oriented)
> then we should do it with the query language were are designing now.
> Otherwise we have to wait until yet another more high level language 
> shows
> up.
>
> If we accept RDF as a means to store knowledge and meaning then an RDF 
> query
> language should be able to handle something like 'relevance' and 'best 
> fit'
>
> Right?
>
> Dirk
>
> ___________________________________
> Dr. Dirk Colaert MD
> Production, Information Systems Architect
> Agfa
> HealthCare Informatics
> call +32 3 444 84 08
> fax  +32 3 444 84 01
>
>

--

Patrick Stickler
Nokia, Finland
patrick.stickler@nokia.com

Received on Sunday, 4 April 2004 03:06:00 UTC