AWS Lex

• Lex is a service for building conversational interfaces using voice and text.

• The most common use-cases of Lex include: Informational bot, Application/Transactional bot, Enterprise Productivity bot and Device Control bot.

• It leverages Lambda for Intent fulfillment, Cognito for user authentication & Polly for text to speech.

• It scales to customers needs and does not impose bandwidth constraints.

• It is a completely managed service so users don’t have to manage scaling of resources or maintenance of code.

• It uses deep learning to improve over time.

• It bot can be created both via Console and REST APIs.

• It provides the option of returning parsed intent and slots back to the client for business logic implementation.

• Users can track metrics for their bot on the ‘Monitoring’ dashboard in the Lex Console.

• It provides SDKs for iOS and Android.

• Users can use AWS Mobile Hub to build, test & monitor bots for their mobile platforms.

• It bots can be published to messaging platforms like Facebook Messenger, Slack, Kik & Twilio SMS.

• Every version of an Amazon Lex bot will have an ARN.

• It supports up to 15 seconds of speech input.

• It supports the following formats for input audio: LPCM and Opus; Supported output audio formats: MPEG, OGG, PCM.

• It can be accessed from VPC via public endpoints for building and running a bot.

• It does not support wake word functionality.

• It provides the ability for users to export their Lex bot schema into a JSON file that is compatible with Amazon Alexa.

• Any content processed by Lex is encrypted and stored at rest in the AWS region where users are using Lex.

• Users can build bots using SDKs: Java, JavaScript, Python, CLI, .NET, Ruby on Rails, PHP, Go & CPP.

• It is supported under Developer Support, Business Support and Enterprise Support plans.

• Every input to Lex bot is counted as a request.

• To build an Lex bot, user will need to identify a set of actions - known as 'intents’.

• To fulfill an intent, the Lex bot needs information from the user. This information is captured in ‘slots’.

• It is capable of eliciting multiple slot values via a multi-turn conversation.