AWS Services for Text/Shape Extraction and Evaluation

AWS can provide you with some services but to accomplish this task you need to handle lots of things AWS services just can support you for this task.

Amazon Textract: Textract can extract printed text (in several languages) from scanned documents, forms, and tables. It uses machine learning to read and process the text from your documents, effectively turning the content into machine-readable data. However, Textract primarily focuses on textual data and has limited support for complex figures, shapes, or handwritten text.

Amazon Rekognition: This service can analyze images and videos for you. You can identify objects, people, text, scenes, activities, as well as any inappropriate content. For your use case, this can be useful to identify and interpret diagrams or shapes if any. Please note that complex recognition might require custom training of the model.

Amazon Comprehend: Once the text is extracted, you can use Comprehend to understand the sentiment or the key phrases in the text. It's an NLP (Natural Language Processing) service that uses machine learning to find insights and relationships in a text.

AWS Lambda: You can use Lambda to create a serverless application that ties together the various AWS services mentioned. AWS Lambda can be triggered when a new document is uploaded to an S3 bucket, process the document using Textract and/or Rekognition, and store the results in a database such as Amazon DynamoDB.

Amazon A2I (Augmented AI): This is useful when you need human review of the ML predictions. If Textract or Rekognition are unable to accurately extract or identify content, A2I can help bridge the gap.

Amazon SageMaker: If your problem is more complex and requires custom models for handwriting recognition or shape detection, SageMaker provides a complete set of tools to train, tune, and deploy your Machine Learning models.