Use this information to configure Intelligence Services.
- Content has metadata automatically applied, and can be categorized or classified based on its business context.
- Unstructured content can be searched and indexed by business context and easily discovered.
- Business rules and processes can automatically be triggered.
Using the Textract OCR solution from Amazon, you can extract plain text from images and PDF files, and then analyze the text. For example, for a given PDF or image, you'll get the raw text from the whole file, tables, forms (using key-value pairs), and check boxes. The extracted data is mapped to properties which are searchable.
- Default configuration
- This option allows you to customize the Request AI renditions action, so that it only calls the renditions that you wish to use. Use these steps if you don't plan to create a custom ML model.
- Custom configuration
- Choose one or more of the following options to create custom ML models:
- Custom entity recognition - configure and deploy a custom AI recognizer. This allows you to identify new entity types that aren't supported by one of the preset entity types.
- Custom document classification - configure and deploy a custom AI classifier. This allows you to classify documents, for example as either an invoice, purchase order, contract, or whatever fits your business model.
- Custom metadata extraction - configure and deploy a custom AI model. This allows you to map basic OCR detected text lines into multi-valued text fields, so they can viewed and searched.
- You can still customize the Request AI renditions action, as in the default configuration.
- It is recommended that you start developing one custom model at a time (i.e. either a recognizer or classifier), and test it thoroughly before adding another.
- Metadata extraction from tables isn't supported.