[Koha-devel] Stemming and zebra

Wed Aug 27 11:48:09 CEST 2014

Hi Mathieu:

I think many of us think certain things happen in Zebra when they actually
happen in Koha before the query ever reaches Zebra ;).

As for stemming, theoretically the language obtained via
“C4::Templates::getlanguage($cgi, 'intranet');” should filter down into the
Snowball stemming. If it isn’t working in French, it might be because the
right locale isn’t being passed to Snowball correctly. That’s very possible
as I think we’re using arbitrary language codes rather than standard locales
in some cases. It looks like there is a fallback to English in
C4::Templates::getlanguage() as well. If it’s not working for French, it
probably just needs a tweak!

Yeah, I first heard about Snowball when reading through Zebra docs, and I
was pleasantly surprised when I saw that Lingua::Stem::Snowball existed as a
Perl interface for the C program.

David Cook

Systems Librarian

Prosentient Systems

72/330 Wattle St, Ultimo, NSW 2007

From: koha-devel-bounces at lists.koha-community.org
[mailto:koha-devel-bounces at lists.koha-community.org] On Behalf Of Mathieu
Saby
Sent: Wednesday, 27 August 2014 7:30 PM
To: koha-devel at lists.koha-community.org
Subject: Re: [Koha-devel] Stemming and zebra

Hi

I had always thought stemming was made by Zebra, and only in english!

In fact the algorithm for french language is here:
http://snowball.tartarus.org/algorithms/french/stemmer.html

(Lingua::Stem::Snowball is a Perl interface to the C version of the Snowball
stemmers)

Mathieu Saby

Le 27/08/2014 10:22, David Cook a écrit :

Hi Francois:

I wrote an email earlier on my tablet, but not 100% sure if it got sent. In
any case, I’m writing again now!

You’ll want to look at C4::Search::_build_stemmed_operand().

Zebra doesn’t actually do any stemming itself. If you read through the Zebra
docs (if you’re masochistic), you’ll notice that they say explicitly that
Zebra doesn’t do any stemming, but that you can do stemming (using a stemmer
like Snowball) while building a query. That’s exactly what we do in Koha.

The Perl module that does the stemming is Lingua::Stem::Snowball. 

However, you might notice that your query’s operands aren’t always stemmed
properly. I haven’t looked in a while, but I think it’s because we don’t
build our queries very well at all (when not using QueryParser). 

If you want to understand why you’re getting “skills” and “fishxsdfe” in
your results, I would suggest running some tests ( using “Data::Dumper” and
“warn” ) so that you can see your query as it is built.

I have a lot of work I want to do on C4::Search::buildQuery, but just don’t
have the time :/.

Unfortunately, at the moment, there is no stemming when using the
QueryParser. However, fortunately, using Lingua::Stem::Snowball with
QueryParser would be really really easy. I think that I’ve written a note on
how to do that somewhere on Bugzilla or maybe on Trello

I hope that helps! Feel free to send me an email or shout at me on IRC if
you want any clarification. I know I probably didn’t make it any clearer but
hopefully this might help you on your path to understanding.

David Cook

Systems Librarian

Prosentient Systems

72/330 Wattle St, Ultimo, NSW 2007

From: koha-devel-bounces at lists.koha-community.org
<mailto:koha-devel-bounces at lists.koha-community.org>
[mailto:koha-devel-bounces at lists.koha-community.org] On Behalf Of Francois
Charbonnier
Sent: Wednesday, 27 August 2014 2:09 AM
To: koha-devel at lists.koha-community.org
<mailto:koha-devel at lists.koha-community.org> 
Subject: [Koha-devel] Stemming and zebra

Hello,

I have tested the QueryStemming system preference on Koha 3.14 (my local
installation) and I'm wondering, does zebra just right truncate the words or
is there an algorithm to find the stems?

I use ICU and I have enabled "QueryWeightFields". I don't have automatic
truncation or fuzzy search on. I use these words for my tests:

&#61623   ski, skiing, skills

&#61623   fish, fished, fishing, fisher, fishxsdfe

Each time, with QueryStemming on, skills and fishxsdfe come out in the
search results. Is it what I should expect? "Skills", maybe but "fishxsdfe"?

Do you know how it works? or have a good example that would help me to
understand?

Thanks!

-- 

François Charbonnier,
Bibl. prof. / Chef de produits

Tél.  : (888) 604-2627
 <mailto:francois.charbonnier at inLibro.com> francois.charbonnier at inLibro.com 

inLibro | pour esprit libre |  <http://www.inLibro.com> www.inLibro.com 

_______________________________________________
Koha-devel mailing list
Koha-devel at lists.koha-community.org
<mailto:Koha-devel at lists.koha-community.org> 
http://lists.koha-community.org/cgi-bin/mailman/listinfo/koha-devel
website : http://www.koha-community.org/
git : http://git.koha-community.org/
bugs : http://bugs.koha-community.org/

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.koha-community.org/pipermail/koha-devel/attachments/20140827/b8503ee5/attachment-0001.html>