[
https://issues.apache.org/jira/browse/OPENNLP-417?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13180183#comment-13180183
]
James Kosin commented on OPENNLP-417:
-------------------------------------
So far, it has improved the performance of the name finder that contains all
types...
Dutch [ned] test data:
Precision: 0.664424218440839
Recall: 0.6418195718654435
F-Measure: 0.6529263076025666
up over 1.5.2 release:
Precision: 0.6509695290858726
Recall: 0.628822629969419
F-Measure: 0.6397044526540929
> Back-to-Back <START><END> tags get improperly set when tagging
> --------------------------------------------------------------
>
> Key: OPENNLP-417
> URL: https://issues.apache.org/jira/browse/OPENNLP-417
> Project: OpenNLP
> Issue Type: Bug
> Components: Name Finder
> Affects Versions: tools-1.5.1-incubating, tools-1.5.2-incubating
> Reporter: James Kosin
> Assignee: James Kosin
> Labels: namefinder, tags, wrong
>
> 1) Special thanks go to Angel Luis Jimenez Martinez for both finding the
> problem, and finding the one line of code that caused this problem.
> This only affects the TokenNameFinder when parsing user input and tagging the
> output.
> The problem was when the name finder model returns the spans, it would
> properly find the Spans and return them; however, it would improperly label
> the Spans when constructing them, if they where back-to-back.
> This can cause the NameFinder to improperly report the wrong tag in this
> situation.
--
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
