Hi all,
I would like to ensure a good coverage and accuracy of the NER annotation on several documents.
The data I’ve created with ner.make-gold
is gold-standard and “complete”, so includes all entity types that occur in the text. I’ve created a model from this data. The data has pretty good coverage of the entity types as well.
Goal: I want to apply the model to a new set of documents and improve the coverage of entity types and accuracy of the model. Surely new documents (quite similar to the model’s articles) con improve coverage so I would like to pass them to annotators.
I have just a couple of questions:
Q1. Apply the model from ner.make-gold
on new documents could be the best-method I can use to get the goal? (model accuracy is over 0.8)
Q2: As an alternative to send annotators entire new documents, how I can save time to annotators sending just the ‘good’ part of text to annotate an gain a better coverage as well ?
Any suggestions or ideas are really appreciated.
All the best
C.