The name finder doesn't filter names based on probability.
All detected names are returned.
Do you train on your own data?
Jörn
On 11/04/2014 04:18 PM, Chris Sardianos wrote:
Hi,
I am using opennlp and its TokenNameFinder and I was wondering if there is a
probability threshold in the finders or models that is used in order to decide
whether an entity is returned or not. My current results have a prob of ~0.86
or more and my goal is to return entities with lower probability. Is there any
option that I can configure in the source files or so?
Thank you in advance!
Chris.
PS: Hope this is the right mailing list to post to. :)
P Please consider the environment before printing this e-mail.
