Passing Ground Truth pdf labelling task output to Comprehend custom entity training

Question

I want to do PDF parsing and identify Custom Named Entities. For that I setup Ground Truth to manually annotate few pdf files following this link - https://docs.aws.amazon.com/comprehend/latest/dg/cer-annotation-pdf.html .

It created a output folder and within that it created another folder with name as  and inside that manifests and annotations folders. When I tried to setup the Comprehend Custom Entity Recognition Job, it asks for bunch of input location. I don't know the values for these fields. From the output in S3 bucket, I came up with the below fields and I don't about what is attribute names.

If possible, Can someone share blog post or some video tutorial on pdf annotation and training similar from PDF input?

1. SageMaker Ground Truth augmented manifest file S3 location
s3://comprehend-semi-structured-docs--/output/labeling-job-20221025T135656/manifests/output/output.manifest

2. S3 prefix for Annotation data files 
s3://comprehend-semi-structured-docs--/output/labeling-job-20221025T135656/annotations

3. S3 prefix for Source documents
s3://comprehend-semi-structured-docs--/src

4.  Attribute names -- ?

Accepted Answer

Some answers were in these 2 blogs

1. [https://aws.amazon.com/blogs/machine-learning/custom-document-annotation-for-extracting-named-entities-in-documents-using-amazon-comprehend/]()

> After you label all the pages, you can find annotations in JSON format in the Amazon S3 location s3://comprehend-semi-structured-docs--/output//annotations/consolidated-annotation/consolidation-response/iteration-1/annotations/

> You can find your output manifest file in the Amazon S3 location s3://comprehend-semi-structured-docs--/output//manifests/output/

2. https://aws.amazon.com/pt/blogs/machine-learning/extract-custom-entities-from-documents-in-their-native-format-with-amazon-comprehend/

> the attribute name would be labeling-job-name-123.

**Comprehend Training Input Location Fields**

**SageMaker Ground Truth augmented manifest file S3 location**

s3://comprehend-semi-structured-docs--/output/resume-labeling-job-20221025T135656/manifests/output/output.manifest

**S3 prefix for Annotation data files**

s3://comprehend-semi-structured-docs--/output/resume-labeling-job-20221025T135656/annotations

**S3 prefix for Source documents**

s3://comprehend-semi-structured-docs--/src

**Attribute names**

will be the Ground Truth job name.

Passing Ground Truth pdf labelling task output to Comprehend custom entity training

相關內容