[
https://issues.apache.org/jira/browse/LUCENE-8777?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16825156#comment-16825156
]
Tomoko Uchida commented on LUCENE-8777:
---------------------------------------
This should be easily fixed by just adding first column of the CSV to
segmentations, when constructing {{UserDictionary}}. And of course tests should
be fixed.
[~cm]: Could you give me some thoughts / comments about current Tokenizer
behaviour (and my proposal here)?
> Inconsistent behavior in JapaneseTokenizer search mode
> ------------------------------------------------------
>
> Key: LUCENE-8777
> URL: https://issues.apache.org/jira/browse/LUCENE-8777
> Project: Lucene - Core
> Issue Type: Improvement
> Components: modules/analysis
> Reporter: Tomoko Uchida
> Priority: Minor
>
> A user reported to me about inconsistent behaviour in JapaneseTokenizer's
> search mode.
> Without user dictionary, JapaneseTokenizer (mode=search) outputs "long token"
> and all of "short (custom segmented) token"s.
> e.g.:
> 関西国際空港 => 関西 / 関西国際空港 / 国際 / 空港
> With user dictionary, JapaneseTokenizer (mode=search) outputs all short
> tokens but not long token.
> e.g.:
> {code}
> $ cat config/userdict.txt
> 関西国際空港,関西 国際 空港,カンサイ コクサイ クウコウ,カスタム名詞
> {code}
> 関西国際空港 => 関西 / 国際 / 空港
>
> This behaviour is confusing for users and would be better to be fixed. I am
> not sure which behaviour is correct, but in my perspective the first one
> (without user dictionary) is preferable.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
