Does OpenNLP use WordNet under the hood for the Named Entity Recognition

Question

I have tried using OpenNLP Tools 1.5 from SourceForge for getting the Named Entites from a text. I did find a JWNL file in the OpenNLP download. Does that imply that OpenNLP in turn is using WordNet for the Named Entity Recognition. (Specifically, does it mean that the Name Model files were generated with WordNet) ?

Answer 1

No, OpenNLP only uses WordNet for co-reference resolution, such as correlating "He" to "Steve Jones" in separate sentences.

WordNet is not used in Named Entity Recognition. The named entity training files are manually annotated fields from various conferences/tests, such as CoNLL.

Source: http://opennlp.sourceforge.net/README.html

Answer 2

As I know OpenNLP use maximum entropy package to provide all their statistical models - pos tagging, sentence detection, NER even tokenization. Wordnet integration is not part of OpenNLP so I think its features were not used in training models.