[
https://issues.apache.org/jira/browse/LUCENE-7420?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15427968#comment-15427968
]
Robert Muir commented on LUCENE-7420:
-------------------------------------
None of these stemmers are 100%. We simply implement the logic here:
http://members.unine.ch/jacques.savoy/clef/
Have you done some more formal analysis/measurement of this change? I am
worried about overstemming errors. I'm not sure we should change savoy's
algorithms without going thru that kind of effort.
> SpanishLightStemmer stemming errors
> -----------------------------------
>
> Key: LUCENE-7420
> URL: https://issues.apache.org/jira/browse/LUCENE-7420
> Project: Lucene - Core
> Issue Type: Bug
> Components: modules/analysis
> Affects Versions: 5.x, 6.x
> Reporter: J Pardos
> Priority: Minor
> Labels: analysis, stemming
> Attachments: SpanishLightStemmer.java.patch
>
>
> SpanishLightStemmer only applies stemming if the original word is 5
> characters or longer. It tries to removing gender and numbers, removing final
> vowel and s if they exist.
> So, perro, perra, perros and perras (dog, bitch, dogs and bitches) all become
> 'perr'.
> The problem arises with shorter words. Gatos and Gatas (cats / female cats)
> all become 'gat', but the singular forms (gato, gata) are below the 5
> character threshold, so they wouldn't be stemmed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
