[openbiblio-dev] Virtuoso and indexes ... can we make use of them?

[William Waites]

* [2010-11-23 12:40:47 +0000] Ben O'Steen <bosteen at gmail.com> écrit:

] Will,
] 
] As virtuoso is running upon a RDB of some sort and you've commented that
] it took many days to build its indexes, is there any way to make use of
] them, via a query or SPARQL?

Whenever you use the bif:contains operator (in a filter or as a
pseudo-predicate) it will use the index. This is how the search is
implemented as well.

] I'm not sure what additions the backend adds for looser text matching of
] literals, but it would be very useful to be able to have inexact
] matching of literals via queries to the triplestore.

There is the possibility of doing something like searching for Edin*
but with two caveats:

   * there must be at least three characters before the *
   * no tokenisation and OR'ing of tokens is done so order matters,
     e.g. "Rom* Juliet" will match "Romeo and Juliet" but "Romeo
     Juliet" will only find results with & (since punctuation is
     apparently stripped by the FTI

This is an area that can probably be improved by experimenting with
different SPARQL queries and perhaps working with OpenLink if we find
something that is difficult to do efficiently.

Cheers,
-w

-- 
William Waites
http://eris.okfn.org/ww/foaf#i
9C7E F636 52F6 1004 E40A  E565 98E3 BBF3 8320 7664