Amazon today announced minor updates, using machine learning, to Textract, its service that extracts written text and other data from documents, as well as tables and forms. As of today, besides files typed in Spanish, Portuguese, French, German, and Italian, Textract now supports handwriting in English documents.
Amazon correctly states that many documents contain a mixture of handwritten and typed text, such as medical intake forms or job applications. Although rivals such as Google and Amazon have for some time offered handwriting recognition-as-a-service, Amazon claims customer demands prompted the launch of its own solution, which operates in tables and forms with both free-form text and text embedded.
For enhanced efficiency, Amazon Web Services (AWS) clients may use the Textract handwriting recognition function in conjunction with Amazon’s Augmented AI (A2I). A2I allows users, either employees or AWS Marketplace contractors, to create workflows for human analysis of the predictions of the machine learning framework. You can upload documents to the Amazon Textract console or send them using AWS Command Line Interface or AWS development software kits.
Aside from new features, Textract remains less holistic than Google’s recently released Document AI (DocAI) platform, a Google Cloud-hosted document processing console. It encourages the creation and adaptation of processing workflows designed with a predefined taxonomy without the need for additional data mapping or training to be carried out. For domain-specific documents, DocAI provides general processors, including a type parser, W9 parser, optical character recognition, document splitter, and custom workflows. Parsers may identify data in documents such as addresses, account numbers, and signatures, as well as extract data such as names of vendors, invoice dates, and terms of payment.
According to some figures, businesses pay an average of $20 to file and store a single document, and just 18 percent of firms consider themselves paperless. An IDC study showed that a 21.3 percent productivity loss accounts for document-related problems, and U.S. firms spend a combined $8 billion per year handling paperwork. This has given rise to a cottage industry, with startups such as Rossum, Anvil, PandaDoc, and others vying for document processing accuracy and pricing, in addition to AI-powered products from tech giants such as Amazon and Google.
Amazon earlier this week also revealed updates to Amazon Lex and Amazon Polly. French, Spanish, Italian, and Canadian French are newly sponsored by Lex, a forum for creating conversational interfaces. Polly, which transforms text into lifelike speech, now features the first Australian English voice from Amazon, Olivia, synthesized with the same neural text-to-speech technology that drives the Brand Voice service from Amazon.