Clarification on Using ner.correct for Pre-Identifying Entities in Data Annotation

Hi, I hope you are doing well.

I would like a custom NER model to aid in data annotation tasks (i.e., I want my custom NER model to pre-identify entities, so the human data annotator either accepts or updates such identified entities). What I have found so far is ner.correct , but I am not sure if:

  • 1) this is the correct match for what I am looking for?
  • 2) this command gets similar results than ner.manual (i.e. a database with the texts, the entities and so on)?

Your guidance on this would be greatly appreciated. Thank you!

Best regards,


1 Like

Hi @agustinadinamarca,

That's right ner.correct is the right choice for this use case.
The output will be almost exactly the same as in the case of ner.manual . The only difference is that the spans annotated by the model that you will accept without correction will have some extra attributes such as source to indicate the model and input_hash to indicate the input the resulted in this annotation. Otherwise the structure will be the same.

1 Like