Understanding document classification with the Amazon Comprehend custom classifier

Sometimes, we receive many documents as a single package, and we need to process each document individually to derive insights from it as per business requirements. To achieve this, one of the major tasks is to categorize and index different types of documents. This later helps in the accurate extraction of information that meets business-specific requirements. This process of categorizing documents is known as document classification.

Let’s look into a claims processing use case in the insurance industry. Claims processing is very much a transactional use case, with millions of claims getting processed every single year in the United States. A manual submission of a claims package is a combination of multiple documents, such as an insurance form, receipts, invoices, ID documents, and some unstructured documents such as doctor’s notes and discharge summaries. A package is a combination...