Hi,
The software I'm building currently makes use of Lucene's MoreLikeThis [1] 
facilities in order to match users' searches with Nodes in a JCR repository. 
MoreLikeThis gives us much better results than, for example, simple full text 
search.

We have currently implemented this by having our own Lucene indices. For 
efficiency and simplicity, we'd prefer to instead use the copy of Lucene that 
is already built into Oak. To do so, I think Oak would need to have an indexer 
configuration for this. Oak would also need a way to pass such queries through 
the jcr query api. I talked to Michael Marth about this: he suggested that we 
would need something similar to "rep:similar". I believe the current 
rep:similar can only take a node to be matched, as opposed to text to be 
matched, so it isn't exactly what we need. It is very close, though.

Any thoughts? I filed [2] for tracking.

Thanks,
Laurie

[1]: 
http://lucene.apache.org/core/3_0_3/api/contrib-queries/org/apache/lucene/search/similar/MoreLikeThis.html
[2]: https://issues.apache.org/jira/browse/OAK-1286

Reply via email to