[
https://issues.apache.org/jira/browse/JENA-1488?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16392865#comment-16392865
]
Bruno P. Kinoshita commented on JENA-1488:
------------------------------------------
Updated my current branch, removing the assembler changes and the analyzer. Now
it actually holds one single file,
[/org/apache/jena/query/text/filter/SelectiveFoldingFilter.java|https://github.com/apache/jena/compare/apache:46e2f56...kinow:d90ffa0]
I have not added tests, not squashed commits, removed main method, etc, as the
code may still need some further massaging.
The output of the main method now would be:
{noformat}
TERM = Senora
TERM = Siobhan
TERM = look
TERM = at
TERM = that
TERM = façade
{noformat}
So the _façade_ keep the cedilla, as it was whitelisted. If the letter 'ñ' was
added to the white-list, then the first term found would actually be _Señora_.
After using the white-list, the code delegates it to a method from the existing
ASCIIFoldingFilter.
Now just need to find a way to rig it up together with Jena text analyzers. I
liked [~code-ferret], though I am not entirely sure where/how to update the
ConfigurableAnalyzer. I tried using it, and noticed I couldn't pass the
white-list when creating an analyzer/filter.
> SelectiveFoldingFilter for jena-text
> ------------------------------------
>
> Key: JENA-1488
> URL: https://issues.apache.org/jira/browse/JENA-1488
> Project: Apache Jena
> Issue Type: Improvement
> Components: Text
> Affects Versions: Jena 3.6.0
> Reporter: Osma Suominen
> Assignee: Bruno P. Kinoshita
> Priority: Major
>
> Currently there's some support for accent folding in jena-text, because
> Lucene provides an ASCIIFoldingFilter. When this filter is enabled, a search
> for "deja vu" will match the literal "déjà vu" in the data.
> But we can't use it here at the National Library of Finland (for Finto.fi /
> Skosmos), because it folds too much! In the Finnish alphabet, in addition to
> the Latin a-z (which are in ASCII) we use the letters åäö and these should
> not be folded to ASCII. So we need a Lucene analyzer that can be configured
> with an exclude list, something like
>
> new SelectiveFoldingFilter(String excludeChars)
>
> and that can be also be configured via the Jena assembler just like other
> analyzers supported by jena-text.
>
> This was also briefly discussed on the skosmos-users mailing list:
> [https://groups.google.com/d/msg/skosmos-users/x3zR_uRBQT0/Q90-O_iDAQAJ]
> Apparently Norwegians have the same problem...
> I've discussed this with [~kinow] and he has some initial code to implement
> this feature, so I think we can turn this into a PR fairly soon.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
