Amazon Lex

Amazon Lex is a service from Amazon Web Services for building conversational interfaces — chatbots and voice bots — using the same automatic speech recognition and natural language understanding technology behind Alexa.

AI Practitioner focus

  • Lex V2 builds multi-turn voice/text bots using automatic speech recognition and natural-language understanding around intents, utterances, slots, sessions, and fulfillment.
  • Choose Lex for a structured conversational interface; choose a foundation-model assistant when broad generative reasoning is required, and still constrain external actions through deterministic authorization.

Key points

  • Handles both text and voice conversations, converting speech to text and then determining the user’s intent from it.
  • Built around intents (what the user wants), utterances (the phrasings that signal an intent) and slots (the parameters a fulfilment step needs).
  • Fulfils intents by invoking AWS Lambda functions, which is how a bot actually looks something up or completes a transaction.
  • Commonly paired with Amazon Connect to automate the front line of a contact centre before escalating to a human agent.
  • Requires no machine learning expertise — a pre-trained service rather than a model you build in Amazon Sage Maker.
  • The boundary that matters: Lex builds the conversation. Amazon Polly only speaks text aloud, Amazon Transcribe only converts speech to text, and Amazon Kendra answers search queries over a document corpus rather than holding a dialogue.

Sources