We had stemming on for authors at first (maybe was the VUFind default way
back when?) and turned it off as soon as we noticed. The initial complaint
was that searching on Rowles gave records for Rowling. and of course
it's not hard to find other examples, esp. with the -ing suffix.
On Mon, Jun
Does Solr support Soundex? (Soundex was originally developed to
assist with alternate spellings of names)
Keith
On Mon, Jun 13, 2011 at 8:08 PM, Jonathan Rochkind rochk...@jhu.edu wrote:
In a Solr-based search, stemming is done at indexing time, into fields with
stemmed tokens.
It seems
On Jun 14, 2011, at 08:10 , Keith Jenkins wrote:
Does Solr support Soundex? (Soundex was originally developed to
assist with alternate spellings of names)
Indeed. And several other phonetic algorithms:
That's an interesting idea, I might try creating author fields with
Soundex normalization rather than the standard English language
'stemming' normalization.
Still curious to get more feedback on what others have done, even if you
didn't consider it carefully, if you're doing it in production
Hey Erik, in that wiki documentation the example it gives is:
filter class=solr.PhoneticFilterFactory encoder=DoubleMetaphone
inject=true/
Do you know what that 'inject' argument is about, and where (if
anywhere) I'd find it (and other available arguments for
PhoneticFilterFactory, which
It's documented in that wiki page link below as true/false -- true will add
tokens to the stream, false will replace the existing token
So if you index cat and it the phonetic filter turns it into KT, it can
either index cat and KT or just KT.
Erik
On Jun 14, 2011, at 10:45 ,
In a Solr-based search, stemming is done at indexing time, into fields with
stemmed tokens.
It seems typical in library-catalog type applications based on Solr to have the
default (or even only) searches be over these stemmed fields, thus
'auto-stemming' to the user. (Search for 'monkey', find