[
https://issues.apache.org/jira/browse/LUCENE-6138?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14261242#comment-14261242
]
Robert Muir commented on LUCENE-6138:
-------------------------------------
One nice thing about the porter stemmer is that it does some basic syllable
counting instead of using arbitrary limits of characters like the light
stemmers do. This is more complicated than checking a limit, but probably the
best solution for this problem. Its probably easier to implement this logic
with the snowball language itself (for this language, the logic already exists
there), but could be done in java too. However these kind of changes need real
evaluation work too. Currently we don't invent these light stemmers, we just
use other people's work.
> ItalianLightStemmer doesn't apply on words shorter then 6 chars in length
> -------------------------------------------------------------------------
>
> Key: LUCENE-6138
> URL: https://issues.apache.org/jira/browse/LUCENE-6138
> Project: Lucene - Core
> Issue Type: Bug
> Components: modules/analysis
> Affects Versions: 4.10.2
> Reporter: Massimo Pasquini
> Priority: Minor
>
> I expect a stemmer to transform nouns in their singular and plural forms into
> a shorter common form. The implementation of the ItalianLightStemmer doesn't
> apply any stemming to words shorter then 6 characters in length. This leads
> to some annoying results:
> singular form | plural form
> 4|5 chars in length (no stemming)
> alga -> alga | alghe -> alghe
> fuga -> fuga | fughe -> fughe
> lega -> lega | leghe -> leghe
> 5|6 chars in length (stemming only on plural form)
> vanga -> vanga | vanghe -> vang
> verga -> verga | verghe -> verg
> I suppose that such limitation on words length is to avoid other side effects
> on shorter words not in the set above, but I think something must be reviewed
> in the code for better results.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
