- From: Johnson, Matthew C. (LNG-HBE) <Matthew.C.Johnson@lexisnexis.com>
- Date: Thu, 22 May 2008 08:42:31 -0400
- To: "Peter Ansell" <ansell.peter@gmail.com>
- Cc: <semantic-web@w3.org>
This discussion is making me think that using SPARQL to [help] answer the question as to whether a given pub links to another might be the right way to go. This brings me back around to questions on the open-world/closed-world assumption. But I just put that question in a response to Alan Ruttenberg's email so I won't ask it again here. > -----Original Message----- > From: Peter Ansell [mailto:ansell.peter@gmail.com] > Sent: Thursday, May 22, 2008 12:14 AM > To: Johnson, Matthew C. (LNG-HBE) > Cc: semantic-web@w3.org > Subject: Re: stating that something doesn't exist > > 2008/5/22 Johnson, Matthew C. (LNG-HBE) > <Matthew.C.Johnson@lexisnexis.com>: > > Hmm. This is interesting, thanks. As far as generating a lot of > > triples from a sparsely linking set of publications, I think my actual > > use-case would probably eliminate that. I suspect that this type of > > query would normally be done on a publication-by-publication basis. So > > I would really be checking if any links exist specifically for > > publication 2 to [specifically] publication 3. This would dramatically > > reduce the number of triples generated (although it might still take > > some time). > > You can modify this query to specifically insert the two publications > into the query and it should dramatically improve the performance over > the all vs all publication because it is still an existence query so > if the SPARQL engine can utilise a database index against specific > publications it should have similar performance to a direct SQL query. > > > Also, I suspect another alternative would be to let the application > > itself do some of the rationalization by simply grabbing all of the > > publications that "pub 2" does link to (a simple query) and then > > checking to see if "pub 3" is in that list. Ideally, I would like the > > data or query to say this directly (really, I'd just like a yes or no > > answer) but if performance becomes an issue, it might be a second > > choice. > > I don't think the query would be too inefficient if you know what the > publications are. I was referring to the all vs all capabilities when > I said it might not be efficient. Basically if there is a row returned > then you have a result, or conversely if no rows are returned then you > have the opposite result. > > Peter
Received on Thursday, 22 May 2008 12:43:16 UTC