[Koha-devel] spell suggestion in rel_3_0 (for kados 1st)

[Joshua Ferraro]

On Fri, Dec 15, 2006 at 05:33:39PM +0100, Paul POULAIN wrote:
> in opac-search.pl, rel_3_0, there is a spelling suggestion when the 
> search don't return any answer.
> It is based on some hardcoded database :
>               DBI->connect( "DBI:mysql:suggest:66.213.78.76",
>                 "auth", "Free2cirC" );
Hehe, so it is. LibLime should be releasing a spellcheck service
based on the API we wrote for English Koha catalogs soon -- it's
one of those projects that's on the back burner. The data we've
been collecting over the past few years has some pretty interesting
uses, beyond Koha even. So I apologize for committing that LL-specific
code. I'll fix it asap.

> I found : make_spellcheck_suggest.pl, in misc directory.
> 
> This script seems to have some problems :
> - it relies on marc_subfield_table & marc_word, that don't exist anymore
> - something is missing at line 97 : %tables variable is missing some 
> tables descriptions (& a "; at the end...)
> - it works with soundex algorythm, that is very poor in french (it 
> doesn't take care of accents & other french-specific things)
Yep, I agree, it's not an ideal method for French and other libraries,
though works quite well for English given a good data set.

> opac-search.pl itself seems to have problems too :
> - $koha_spsuggest_insert is never called : is it a feature you have 
> abandonned ?
Yep, it's abandoned.

> I was wondering :
> - if there is a more uptodate script that you forgot to commit, because 
> this one can't work (I can't find the script in dev_week) ?
Probably, I'll check.

> - if we could use a specific soundex, depending on the language : there 
> are some well working soundex_like algorythm for french & I bet other 
> languages have their. That would require removing the 
> str_cmp(soundex(suggestion),soundex(?)) from opac-search
Yea, the whole thing needs to be revised and generalized with some new
sysprefs for changing the settings, etc. I'll have a go at it this
weekend.

> - if we could use Aspell for spellchecking. I tried to investigate this 
> option, it works very very well with common names, and we mostly want it 
> to work for proper names like authors, so I think it's not something to 
> investigate too deeply.
Agreed.

> - who will be the next french president, in 2007, May ?
Hehe, well I hope I know the answer to who will not be the next US
president ;-).

-- 
Joshua Ferraro                       SUPPORT FOR OPEN-SOURCE SOFTWARE
President, Technology       migration, training, maintenance, support
LibLime                                Featuring Koha Open-Source ILS
jmf at liblime.com |Full Demos at http://liblime.com/koha |1(888)KohaILS