Custom entity annotation with groundtruth

Question

I'm trying to train a custom entity recognizer with aws comprehend by using annotations labelled with aws groundtruth.   
  
Via ground truth, annotations are stored in the augmented manifest file in s3 but the format of the annotations doesn't look easily compatible with the expected format of the annotations to train a custom entity recognizer using comprehend.  
  
Is there an easy way to combine these two tools or would I need to manipulate the annotated data in a more manual way to match the expected training data format for comprehend?

AWS-User-8841074 · Answer

Please refer to this solution to train Comprehend Custom entity recognizers using data annotated with Groundtruth - https://aws.amazon.com/blogs/machine-learning/developing-ner-models-with-amazon-sagemaker-ground-truth-and-amazon-comprehend/

Garyjb · Answer

thank you!   
  
A code snippet that might help the process would be really useful here if you're able to help?

AWS-User-8841074 · Answer

Thank you for your question.  
  
Comprehend currently does not support training of a customer entity recognizer using annotations labelled with Groundtruth in the exact output format of Groundtruth. We have submitted a feature request on your behalf to support this.   
  
In the mean time unfortunately, you will have to write some custom code to transform the output of Groundtruth to the one supported by Comprehend Entity Recognizer training.  
  
Please let us know if you have issues getting this done and we can get back with a code snippet that will help accomplish this.  
  
Thanks,  
Seema

Custom entity annotation with groundtruth

Relevant content