Amazon Textract

I like Textract's ability to quickly and reliably extract structured and unstructured data from a wide variety of documents including pngs and pdfs. I value Textracts ability to automatically extract text, tables, key fields at scale without buildling or maintaining OCT infrastructure.

Pre-made document processing types.

1. It has a solid baseline Optical Character Recognition, especially for structured PDFs and standard forms. 2. It's good at table extraction, which is how most of the information is presented when it comes to policy documents for insurance. 3. It is easy to integrate since the API is straightforward and fits nicely into an already existing AWS infrastructure.

Costs can rise with large volumes of documents or high frequency processing. Some formatting or nested tables may not always extract perfectly on the first pass. Some of the extracted fields need additional validation or post-processing logic.

It does not support Cyrillic characters.

1. It requires a lot of post-processing in order to make the output usable. 2. Industry-specific terminology isn't well understood, leading to a lot of wrong labelings. 3. Accuracy drops on complex layouts.