[
https://issues.apache.org/jira/browse/LUCENE-3475?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16357375#comment-16357375
]
Adrien Grand commented on LUCENE-3475:
--------------------------------------
Exciting! I'll need some time to digest this new implementation but I already
have some questions/remarks :)
- Is my understanding correct that this filter can consume graphs but won't
generate a graph unless the input is a graph too (ie. it won't require
FlattenGraphTokenFilter at index time unless the input is a graph)?
- At first sight, this impl looks compatible with the existing ShingleFilter,
or do I miss something? Should we just replace it?
- Can the tests create token stream graphs explicitly with CannedTokenStream
instead of implicitly with SynonymGraphTokenStream?
- I'd be curious to see tests on more complex token stream graphs eg. like
those used in TestFlattenGraphFilter.testWTF1/testWTF2, does your impl support
such complex graphs?
> ShingleFilter should handle positionIncrement of zero, e.g. synonyms
> --------------------------------------------------------------------
>
> Key: LUCENE-3475
> URL: https://issues.apache.org/jira/browse/LUCENE-3475
> Project: Lucene - Core
> Issue Type: New Feature
> Components: modules/analysis
> Reporter: Cameron
> Assignee: Alan Woodward
> Priority: Minor
> Labels: newdev
> Attachments: LUCENE-3475.patch
>
>
> ShingleFilter is creating shingles for a single term that has been expanded
> by synonyms when it shouldn't. The position increment is 0.
> As an example, I have an Analyzer with a SynonymFilter followed by a
> ShingleFilter. Assuming car and auto are synonyms, the SynonymFilter produces
> two tokens and position 1: car, auto. The ShingleFilter is then producing 3
> tokens, when there should only be two: car, car auto, auto. This behavior
> seems incorrect.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
