A dead-simple annotation app to supervise the predictions of your NER algorithms. Upload a .csv-file that contains the strings, along with the predicted start- and end-characters of the Named Entities, as well as their type. This app will then provide an easy interface to accept or reject those predictions, and bulk-download the results.
Inspired by Prodigy from the spaCy team, without all the fancy-ness and features.
Shouldn't I just build a good training set instead?
Sure, but once your algorithm reaches halfway decent predictions, it'll be faster to just tell it what it gets right or wrong, versus more tediously continuing to expand a valid training set.
This becomes more obvous when you have a model that supports some kind of inline training and an API that this app could connect to directly, instead of doing .csv im/exports.