[
https://issues.apache.org/jira/browse/OPENNLP-428?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13204540#comment-13204540
]
Katrin Tomanek commented on OPENNLP-428:
----------------------------------------
Short explanation to my patch: the sentence detector logic now has a new
parameter, which can be specified for the CMD-Line tools (eosChar). Here, you
can specify characters to be used as EOS symbols. eosChars is optinal; if not
specified, the default EOS symbols (as defined in the Factory,
language-dependent) are used.
Note: specifying eosChars will override the logic caused by language in the
Factory class.
I also modified the cross-validator, the evaluator, the trainier etc to use the
EOS symbols.
EOS symbols, if provided, are stored in the models manifest.
> make EOS character set configurable
> -----------------------------------
>
> Key: OPENNLP-428
> URL: https://issues.apache.org/jira/browse/OPENNLP-428
> Project: OpenNLP
> Issue Type: Improvement
> Components: Sentence Detector
> Reporter: Katrin Tomanek
> Priority: Minor
> Fix For: tools-1.5.3-incubating
>
> Attachments: patch_eos_characters
>
>
> Currently, the EOS symbols to be used by the sentence detector cannot be
> configured (at the moment, a user would have to make changes in
> opennlp.tools.sentdetect.lang.Factory
> Since it is important to use the same EOS symbols during training and during
> testing/prediction, the EOS symbols should be stored with the model's
> properties
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
