Besides chatbots, Lex can also be used to build web and mobile apps for human-machine interactions such as order taking, information provision, work streamlining, and as a control mechanism for robots, toys and drones, AWS said.
“The service understands context. That allows you to build sophisticated, multi-step conversational interfaces where you can continue to clarify and guide the user intuitively to what they want to go and fulfil,” general manager and director of AWS Dr Matt Wood told the AWS:Invent conference.
The service only works with Facebook Messenger at the moment, with Slack and Twilio support expected to arrive soon.
Lex also works with AWS services such as Lambda for business logic, Mobile Hub for conversational interfaces to enterprise data, and the Cloudwatch monitor.
It uses the automatic speech recognition and natural language understanding technologies that Alexa provides.
AWS said Lex will be a fully managed service with no set-up, management or infrastructure scaling needed.
Lex is available as a wait-listed preview in the US East (North Virginia) AWS region only. AWS did not say when it will be publicly available or when it will reach other regions.
Lex is free for up to 10,000 text and 5000 speech requests per month for the first year of use. It will then cost US$4 per 1000 speech and US$0.75 per 1000 text requests.
AWS chief executive Andy Jassy also touted two further artificial intelligence services: the first, Amazon Recognition, lets app developers pass images through APIs or software development kits to the cloud, where the service will detect objects in the picture.
This could include people or inanimate objects, facial matching for security, and more. Jassy said Recognition can identify sentiment such as whether a person is smiling or frowning.
Jassy also announced the Polly text-to-speech deep learning service, which allows developers to send streams of text and quickly get an MP3 audio file in return, with life-like voices.
Polly is designed to be responsive and return the audio quickly so as to reduce perceived delays in conversational flows for more natural discussions.
The service currently provides 47 voices in 24 languages, with male and female speakers and a variety of accents.
Paris Cowan attended Re:Invent as a guest of AWS