Open GoogleCodeExporter opened 9 years ago
As far as I understand, the combiner is comes into play when multiple NER
models are used and has a resolver strategy which is used when entities overlap
or conflict, keeping only one of them.
DKPro Core also currently does not use SUTime.
Using the NERClassifierCombiner would certainly be a good idea.
We currently assume that one component has one model. However, the
NERClassifierCombiner would need to be fed with multiple models. We need to
find a good way of handling such a case.
Optimally, the StanfordNamedEntityRecognizer integration in DKPro Core should
behave in the very same way as the entity recognizer in a CoreNLP pipeline.
If you have any suggestions for us to improve the present situation, please
shoot.
Original comment by richard.eckart
on 20 Oct 2014 at 8:14
Original issue reported on code.google.com by
jkirschn...@gmail.com
on 20 Oct 2014 at 5:04