Aldo Bucchi wrote:
Hi Ivan,
On Sat, Jul 11, 2009 at 1:39 AM, Ivan
Mikhailov<imikhai...@openlinksw.com> wrote:
Hello Aldo,
I really like this idea. Say, when search for standard, first search in
more important international documents, then common national, then
specific for the specific application, then local recs. That sounds
Yeah, that's one possible use case.
There is also another one, very common:
Say my language preferences are English, Spanish, French, Italian, in
that order. I want my user agent to know this when looking for a
suitable label or text literal and fallback accordingly.
We already do this for DBpedia via a combination of rdfs:label (some owl
rules) and TCN.
>From an implementation standpoint, we can easily achieve this now
craftin a SPARQL with a set of Optionals.
However, this generates responses that may carry more than one
alternative. While this may not be important in general, when dealing
with large pieces of text ( like dbpediaProp:abstract ), this can
become a considerable overhead on the network level ( not to any
possible evaluation overhead ).
So I was trying to use union. First alternative: English. Second:
Spanish, etc etc. And then using Limit 1 so as to stop evaluation as
soon as one alternative was matched. But I found no way to control
order.
Rumi: We need to make a technical note / tutorial showing how we use TCN
and our rdfs:labels rules to achieve the above. Check with Carl if we
actually have this
in the Linked Data Deployment tutorials.
Kingsley
nice. Unfortunately, there's no appropriate SQL infrastructure, hence no
chance for quick implementation. The only extension we have for UNION is
so-called BEST EFFORT UNION that lets gather data from multiple
unreliable remote sources without halt on error if some sources are
temporarily unavailable --- the result is formed from data returned by
live instances.
Nice to know this ;)
Anyway I'll bugzilla this idea as an enhancement request to myself, to
not forget.
Cool!
Thanks,
A
Best Regards,
Ivan.
On Fri, 2009-07-10 at 16:08 -0400, Aldo Bucchi wrote:
Hello,
Is it possible to indicate an order of preference for a set of
alternative matches ( unions )?
We have found that explicit order is not respected when querying LOD (
however, it does work against Virtuoso 5 ).
Is this a Virtuoso cluster/anytime introduced behaviour or is this how
SPARQL is supposed to behave?
I have been trying to find some literature on this. I am 90% sure I
once relied on the order of the UNION patterns for something similar.
But I might be very mistaken.
Thanks,
A
--
Regards,
Kingsley Idehen Weblog: http://www.openlinksw.com/blog/~kidehen
President & CEO
OpenLink Software Web: http://www.openlinksw.com
