Amazon Textract now extracts text even more accurately, from even more types of documents
Amazon Textract is a machine learning service that makes it easy to extract text and data from virtually any document. One advantage of services like Textract is that customers benefit from continuous improvement over time. Today, we are pleased to announce a set of quality enhancements that make Amazon Textract even more accurate.
First, we have improved the accuracy our text recognition feature. Second, Amazon Textract now more accurately corrects the rotation of documents and isolates documents from their backgrounds, for more accurate text extraction. These benefits apply to many types of documents, but they are especially pronounced for documents with sparse text, nonstandard paper sizes, small deformations in the paper such as bent corners, extreme or unusual backgrounds surrounding the document, and even for documents that are partially covered. Finally, we have rescaled the confidence scores of our text detection feature to make them better aligned to the underlying accuracy of our models.
You can get started with Amazon Textract today here.