Amazon Textract

2019: Announcement

At the end of May, 2019 Amazon started a cloud service for recognition of documents of Textract which is capable to take automatically from pages the text, tables and other data. Different formats, including JPEG, PNG and PDF are supported.

Textract treats programs of optical text recognition (OCR), as well as, for example, Abbyy FineReader. Unlike many OCR solutions of Textract not just takes the text from  documents,  but also  distinguishes their  format and  contents. For example, it  distinguishes tables and  forms in  documents,  including in  checks, tax declarations and consignment notes  and also supports graphic formats. After recognition of software independently structures data.

Released Amazon the competitor of ABBYY FineReader

Amazon claims that the Textract service is capable to define passport data, dates of birth and  the addresses then it is correct to interpret regardless of  in  what place of the page they are.  In case of change of a template a system  will not pass the wrong result.

According to developers, it was succeeded to achieve high efficiency of recognition due to use of the machine learning (ML) for processing of millions of documents. As a result a system learned to identify correctly the text and objects "practically in any" a document type.

Developers for connection of Textract to the applications do not need to be machine learning specialists, the vice president of department of Amazon Machine Learning Swami Sivasubramanian says. They can take the text and data, using DBMS and analytical services Amazon and to adjust integration with other MO-services.

Textract is intended for automatic recognition of a large number of documents. The cost of use of service begins with $1.5 for 1000 processed pages.^[1]

Notes