parameters for finetunning

[Ashbajawed]

trafficstars

Are there any tunable parameters to improve the performance of ner without retraining?

[labusch]

Hi, no, any change in entity recognition behavior requires a change of the model, i.e., either to retrain it or to replace it by another better model. In particular ground-truth that is specific to your use case helps.

In the mean time (since our 2019 publication) there have been many better models published that could be used as an replacement, for instance: https://github.com/dbmdz/clef-hipe