Posted On: Jun 5, 2020
Amazon Augmented AI (Amazon A2I) is an AWS service that makes it easy to build the workflows required for human review of ML predictions (such as predictions from Amazon Rekognition, Amazon Textract, Amazon Translate or Amazon Comprehend). Amazon A2I provides you the option to work with human reviewers or “workers” inside your own organization through a private workforce, Amazon Mechanical Turk workforce of over 500,000 independent contractors, or vendor-managed workforces pre-screened by AWS for quality and security procedures. Starting today, for your private workforce, Amazon A2I provides additional metadata for each worker that reviews your data, enabling you to uniquely identify them and implement quality control for your workforce.
When you select the option to use your own private workforce, each human review answer uploaded to your Amazon S3 bucket now contains the worker’s unique ID known as “sub”, short for subject. This “sub” id refers to a particular worker from your private workforce (managed for Amazon A2I through Amazon Cognito). This new ability to uniquely identify your workers enables you to have better quality control over your private workforce - you can now analyze your human answers, identify outliers, and conduct training to improve your workforce’s human reviews.
This feature is available in all AWS regions where the Amazon A2I service is: US East 1 (N. Virginia), US East 2 (Ohio), US West 2 (Oregon), Canada Central (Montreal), EU Central (Frankfurt), EU West 1 (Ireland), EU West 2 (London), Asia Pacific (Singapore), Asia Pacific (Tokyo), Asia Pacific (Sydney), Asia Pacific (Seoul) and Asia Pacific (Mumbai).
To get started with Amazon A2I head to the documentation, Using Amazon Augmented AI for Human Review. To learn more about Amazon A2I Private Workforces read Use a Private Workforce. See Monitor and Manage Your Human Loop for details on where to find your human review answers. The ListUsers API provided by Amazon Cognito lets you find workers based on their “sub” using filters. For video presentations, sample Jupyter notebooks or information about use cases like document processing, content moderation, sentiment analysis, text translation, object detection and others see Amazon A2I resources page.