[ckan-dev] Semantic Search - Extension

Mon Dec 3 11:35:16 UTC 2012

This is what I found so far
http://lucene.472066.n3.nabble.com/filter-query-from-external-list-of-Solr-unique-IDs-td1709060.html

To pass a list of IDs to SOLR was my initial idea, too. However, as the  
community there states this approach does have very tight limits (number  
of OR clauses allowed).

What do you recommend?

On Fri, 30 Nov 2012 11:44:48 +0100, Sven R. Kunze  
<sven.kunze at s2007.tu-chemnitz.de> wrote:

> Hi guys,
>
> I was currently thinking about trying to use SOLR for indexing the  
> following data:
> - list of URIs indicating the vocabularies, predicates, classes and  
> entities (for topic search)
> - latitude, longitude, radius (for geo search)
> - min time, max time (for time search)
>
> However, I would face the following issue:
> - how can I assure, that e.g. subclasses of a class will also be found  
> although it is not mention in that RDF dataset explicitly
> => materialization could do that, but that is the business of a  
> triplestore
> - so I could let a triplestore create all additional triples and got  
> them indexed
> - however forward-chaining is not the best way as it causes severe  
> issues when updating (rebuilding the complete closure, indexes etc.)
> => therefore, I'd like to handle the filtering myself via SPARQL queries  
> (where backward chaining can be done)
>
> Implications:
> The issue with the post-filtering (as the extension works now) is that  
> the facets aren't updated correctly.
> So, pre-filtering would be more adequate.
> Is there a way to pass a list of relevant items to SOLR?
>
> The idea is that a triplestore could (or could not defined by the admin)  
> filter out the datasets that match the specified search criteria (topic,  
> geo, time) and SOLR could run its regular search based on that  
> pre-filtered list of datasets.
>
> One more thing: if the described approach worked, could it also be  
> possible to pre-sort the list for SOLR (not by an index value but by a  
> pre-sorting the given list)?
>
> Cheers,
> Sven