Amazon Web Services (AWS), which is hosting its re-Invent conference this week, presented new services that will allow developers to add artificial intelligence (AI), functionality to their apps without requiring any programming or machine learning expertise.
The company announced four new AI offerings at different stages of availability: Amazon Personalize and Amazon Forecast, Amazon Textract, Amazon Comprehend Med, and Amazon Textract.
Amazon Personalize is now available in preview. It’s a recommendation service that uses the same technology that provides product recommendations to Amazon.com customers. Personalize is designed to assist developers with all levels of experience in creating and deploying recommendation models that can be used for a variety of uses, including news articles and digital media.
Julien Simon, an AWS machine learning and AI evangelist, wrote in a blog post, “Using AutoML,” a new process that automates complicated machine learning tasks. Personalize performs and accelerates the hard work required to design and train a machine-learning model.
Personalize can work with streaming data in real-time and data stored in an Amazon S3 bucket. It has built-in recommendation algorithms, which Simon called “recipes” and can be customized by users to suit their needs.
Amazon Forecast is also available in preview. Forecast, like Personalize is built on existing technology in Amazon.com’s online e-commerce business. Forecast uses historical trends and time-dependent data for future forecasting.
Forecast is a managed service. Users no longer need to create their own forecasting models or provision compute resources. Forecast also eliminates the need for users pay upfront for resources they might not use.
“Amazon Forecast does not require any machine learning experience to get going. AWS stated in its product page that historical data is all you need. Amazon Forecast will automatically review your data and identify the most meaningful data. It will then create a forecasting model that can make predictions up to 50% more accurate compared to looking at only time series data.
Amazon Textract is a new natural language capability, also in preview. It scans documents of virtually any type and extracts relevant data using machine-learning. Textract promises to allow users to process millions of pages in hours, without the need for manual coding.
Amazon Comprehend is a Textract-like tool, but it’s for health care. Amazon Comprehend can scan unstructured medical records such as patient records, notes from health providers, and clinical trials. It then extracts the relevant medical information from the documents using machine learning. According to AWS’ product page, this includes “medications and medical conditions, test and treatment and procedures (TTP), anatomy and protected health information [PHI]).
Amazon Comprehend medical builds on the already existing Amazon Comprehend Text-analysis Service, which was presented at last year’s reInvent conference. Amazon Comprehend is able to understand general-purpose text. Simon explained that Amazon Comprehend can understand clinical documents. However, we have been asked by healthcare customers to create a version of Amazon Comprehend that is tailored to their specific needs.
Amazon Comprehend Med is now available in all regions of AWS, including Northern Virginia, Ohio and Oregon.
