I am using Semantic Vectors[1] implementation of LSA in a large scale digital library project called Project Torngat[2]. I presented some of the work at the European Conference on Digital Libraries (ECDL)[3], at the 'Very Large Digital Libraries (VLDL) workshop[4] in September. A pre-print of the paper is here[5]. Badsically I used it to generate similarities of journals using the full-text of all their articles, and used this to create a 2-D map of the journal space.
The Semantic Vectors works very very well, but I believe it is not yet a stable (API or functionality) codebase, more of an experimental-as-yet effort, so it may be a moving target for anyone wanting to integrate it. Gllen Newton http://zzzoot.blogspot.com/ [1]http://code.google.com/p/semanticvectors/ [2]http://lab.cisti-icist.nrc-cnrc.gc.ca/cistilabswiki/index.php/Torngat1 [3]http://www.ecdl2009.eu/ [4]http://www.delos.info/vldl2009/ [5]http://cuvier.cisti.nrc.ca/~gnewton/events/2009/ecdl2009Newton_20090723.pdf 2009/10/30 György Frivolt <fifigy...@gmail.com>: > That's what I found so far too. Seems to be usable. It uses a variant of > LSA, probabilistic LSA. > Writing a module for it.. have to rephrase my question them. Where to start > with writing a module for solr.. Should be useful for diving deeper in solr. > > > > On Fri, Oct 30, 2009 at 9:56 AM, Paul Libbrecht <p...@activemath.org> wrote: > >> Not with solr but with Lucene, there is the project called semanticvectors. >> It would be cute to make it a solr module. >> >> paul >> >> >> Le 30-oct.-09 à 09:17, György Frivolt a écrit : >> >> >> Hi, >>> Does anyone of you have experiences with using LSA, Latent Semantic >>> Analysis with Solr? I would like to search for expressions, but also find >>> records, which have context relevant for the given query. Would appreciate >>> any clue where to start, what to consider. >>> >>> Bw, Georg >>> >> >> > -- -
