tecoholic / ner-annotator

Named Entity Recognition (NER) Annotation tool for SpaCy. Generates Traning Data as a JSON which can be readily used.
https://tecoholic.github.io/ner-annotator/
MIT License
548 stars 161 forks source link

Annotation import process is confusing. #103

Closed arkob closed 2 months ago

arkob commented 9 months ago

Can you add the feature to revise annotations? I have .json file with annotations done by your tool. I want to revise them and check. No option to open .json file and revise it.

alvi-khan commented 9 months ago

image

Hello! It's possible to open an existing annotation file by first opening the base text file and then importing the .json file using the menu option.

arkob commented 9 months ago

Thank you. It works. I was initially confused by the warning message that it would discard all current annotations when I wanted to import it and the text file was open.

alvi-khan commented 9 months ago

I can see how the import process can be confusing. I've had more questions about this too. I'm keeping this issue open until this is improved.

GiovanniNova commented 6 months ago

We should be able to import straight from the .json file since it already contains the text. Why is the .txt file necessary?

alvi-khan commented 6 months ago

We should be able to import straight from the .json file since it already contains the text. Why is the .txt file necessary?

The .txt and .json files aren't exactly the same unfortunately. For example, if you skip a sentence then the .json file has a NULL value in the place of the sentence.

I also disagree that we should import the text from the .json file. In my own experience, a major use case for importing the .json file is to review the annotations done by someone else. You would then be trusting that the .json file really did include all of the sentences from the original text and that the original text was unmodified.

GiovanniNova commented 3 months ago

It would be a nice option thats all, a warning could be displayed when importing the .json without the .txt file.