Amazon Textract announces accuracy and feature updates to DetectDocumentText and AnalyzeDocument APIs

Amazon Textract is a managed machine learning service that automatically extracts text, handwriting, and data from any document or image. We regularly improve the underlying machine learning models based on customer feedback to provide even better accuracy. Today, we are pleased to announce feature and accuracy updates to the text detection model used in Textract DetectDocumentText and AnalyzeDocument APIs.

This update adds support for superscripts, subscripts, and rotated text in documents. The update also includes accuracy improvements for text detection in box forms, extraction of visually similar character sets (e.g., ‘0’ vs. ‘O’), and lower-resolution documents such as faxes.

This update is now available in US East (Ohio, N. Virginia), US West (N. California, Oregon), Asia Pacific (Mumbai, Seoul, Singapore, Sydney), Canada (Central), Europe (Frankfurt, Ireland, London, Paris, Spain), and AWS GovCloud (US-East, US-West) Regions.

To get started, log on to the Amazon Textract console. To learn more about Textract capabilities, please visit the Amazon Textract website, developer guide, or resources page.