Error in Custom Entity Recognition (AWS Comprehend)

Question

Hello,

I'm trying to build a  Custom Entity Recognition model, and I keep getting this error (We found an insufficient number of matches in the training document corpus for the entity Experience. A minimum of 100 matches for each provided entity are required. , exit code: 255)

I tried to increase the training data, then it gives me this error (Custom entity recognition supports a maximum of 120000 samples. , exit code: 255)

Can someone please suggest a solution?

Thanks

Answer

Hello Mohamed,

Amazon Comprehend team has updated the custom NER annotation limitation, from minimum 100 annotation per entity type to 25 annotations per entity type. Check: [Amazon Comprehend announces lower annotation limits for custom entity recognition](https://aws.amazon.com/blogs/machine-learning/amazon-comprehend-announces-lower-annotation-limits-for-custom-entity-recognition/) for more information.

For the better accuracy, when preparing the training data, make sure you have balanced annotation numbers across entity types and followed [Amazon Comprehend Guidelines](https://docs.aws.amazon.com/comprehend/latest/dg/guidelines-and-limits.html).

Answer

Hello,
It looks like that some of  your entity have less then 100 examples and when you add more examples you reached maximum supported samples limit. 
You should analyze your text to understand how may examples you have for each entity. Then you could reduce number of examples for entities with high number of examples and add more for those entities which have less then 100.

Error in Custom Entity Recognition (AWS Comprehend)

関連するコンテンツ