Data entry ops · Production

Multilingual content processing using Amazon Bedrock and Amazon A2I

The problem

Multinational companies receive invoices, contracts, and other documents from various regions in languages such as Arabic, Chinese, Russian, or Hindi that may not be supported by existing document extraction software; handling such complex and sensitive documents requires accuracy, consistency, and compliance, often necessitating human oversight.

Workflow diagram · grounded in source
1
Document acquisition from S3
trigger
“The first stage of the pipeline acquires input documents from Amazon Simple Storage Service (Amazon S3). In this stage, we store initial document information in an Amazon DynamoDB table after receiving an Amazon S3 event notification. We…”
2
Schema-grounded AI extraction
ai_action
“We specifically used the Rhubarb Python framework to extract JSON schema-based data from the documents. Rhubarb is a lightweight Python framework built from the ground up to enable document understanding tasks using multi-modal LLMs. It …”
3
Custom business rules validation
validation
“Custom business rules are applied to the reshaped output containing information about tables in the document. Custom rules might include table format detection (such as detecting that a table contains invoice transactions) or column vali…”
4
Reshape for A2I format
integration
“JSON extracted in the previous step is reshaped in the format supported by Amazon A2I and prepared for augmentation.”
5
Human review and augmentation
human_review
“Human annotators use Amazon A2I to review the document and augment it with any information that was missed. A custom UI built with ReactJS is provided to human reviewers to intuitively and efficiently review and correct issues in the doc…”
6
Catalog validated documents
output
“Documents that pass human review are cataloged into an Excel workbook so your business teams can consume them.”
Reported outcome

The reference solution demonstrates an end-to-end approach for multilingual document ingestion and content extraction, enabling organizations to efficiently process documents in multiple languages and extract relevant insights while incorporating human validation.

Reported stack
Amazon BedrockAmazon A2IClaude V3Amazon Step FunctionsAmazon S3Amazon SageMakerAWS LambdaAmazon SQSAWS CDKAWS CloudFormationRhubarbReactJS
Source
https://aws.amazon.com/blogs/machine-learning/multilingual-content-processing-using-amazon-bedrock-and-amazon-a2i?tag=soumet-20
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The reference solution demonstrates an end-to-end approach for multilingual document ingestion and content extraction, enabling organizations to efficiently process documents in multiple languages and extract relevant…

What tools did this team use?

Amazon Bedrock, Amazon A2I, Claude V3, Amazon Step Functions, Amazon S3, Amazon SageMaker, AWS Lambda, Amazon SQS, AWS CDK, AWS CloudFormation.

How is this data entry ops AI workflow structured?

Document acquisition from S3 → Schema-grounded AI extraction → Custom business rules validation → Reshape for A2I format → Human review and augmentation → Catalog validated documents.

Related data entry ops cases