1
$\begingroup$

I am looking for an NLP annotation tool/library that supports active learning. I am looking for something that works in this scenario:

  • Annotating N samples.

  • Training a model on the annotated data.

  • Getting the model's predictions on the next N unlabeled data.

  • Correct/annotate (manually by the annotator) the annotations of the unlabeled data.

  • Retrain the model by including the labeled data from the last step.

I found a library that's called Prodigy but it's not free. Any suggestion for a free library?

$\endgroup$

2 Answers 2

2
$\begingroup$

For my projects I use NLP Lab, by John Snow Labs. It provides automated annotation and model training, saves time compared to other tools, and is completely free of charge. Another impressive thing about it (which I found) is that you don't need to have a prior experience in coding, because there is no coding involved. You can even invite your team members and collaborate with them in your project.

The documentation can be a long read, I have provided a direct link here: docs

$\endgroup$
1
$\begingroup$

There is a tool Acharya which does this, available here (https://github.com/astutic/Acharya).

You can upload your dataset and then add any algo to train a custom model and use it for further annotations.

Their documentation regarding adding an algorithm is a bit confusing but adding it is pretty straight forward, to begin with you can simply copy paste a configuration template which they provide into the Add algo section, this config template which trains a spacy ner model and is available here:

https://github.com/astutic/acharya-spacy/blob/setup/algo_configuration.yaml

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.