Textract fails to extract correct values

Question

We are using AWS Textract modules Analyse Document and Analyse Expenses to extract data from Invoices and Purchase Orders. We manually extract fields which are failed in auto extract but for similar such documents, still those fields are not getting extracted.

Please find attached screen shots of the document where some fields are not getting extracted (Bill To, Customer Name, Address are randomly extracted) Table line items not at all getting extracted for page 2.  I have also attached the screenshots of JSON data. 
For confidentiality of the data I have blurred the crucial information.
![BILL-TO not getting extracted](/media/postImages/original/IMtMZ605nZSbmGWpT2jefRAA)
![BILL-TO not getting extracted](/media/postImages/original/IMN2nsNWnrR8iJWGMEdU3Stg)
![Table with data on Page 2 not extracted](/media/postImages/original/IM4i6vvjo8QTiSQGyhfTN7EQ)
![Table with data on Page 2 not extracted](/media/postImages/original/IMngk8UEVqR3emRulowb_efg)
![Table with wrong values extraction](/media/postImages/original/IM8eqSlJg4QD-xFqbsTdj8Eg)
![Table with wrong values extraction](/media/postImages/original/IMOy2P8tHfTRuAyPIrFq_jtg)
How can we fix this? Please help us on resolving these issues as it is not giving confidence to customer.

Alan_L · Answer

Thank you for using Amazon Textract. As a managed machine learning service, we are continuously improving the quality of our models and releasing new features. In order to help us improve our models for your documents, please open a customer support ticket and share details to help us analyze further. Additionally, please look out for announcements regarding our model quality updates and new feature announcements that are announced on the AWS Textract what’s new post channel.

Larzzon · Answer

I have a similar issue!

Textract fails to extract correct values

Relevant content