Ref: https://learn.cantrill.io/courses/1820301/lectures/42177845

• Summary:
• Contents:

Amazon Lex - Key Concepts

• 🔧 Interactive chatbots
  • Text or voice conversational interfaces
  • ❗Powers Amazon Alexa!
  • Two functionalities:
    1. Automatic Speech Recognition (ASR)
    2. Natural Language Understanding (NLU)
       • Lex discovers the user's intent(s)
• Features
  • Scales well, quick to deploy
  • Pay-as-you-go pricing → great for event-driven architectures
  • Integrates with other services (e.g. Connect)
    • Lex is not a service that you will often use from Console UI, often baked in with other services
    • Builds understanding into your applications → you don't need to program that functionality yourself
• Use cases: all sorts of voice bots
  • Chatbots
  • Voice Assistants
  • Q&A Bots
  • Info/Enterprise bots

Amazon Lex - NLU Concepts

• Intent: an action the user wants to perform
  • e.g. ordering a hamburger
  • Intents often fulfilled with Lambda integration
  • Chain of intents: “order a hamburger” → “put extra cheese” → “without onion”…
    • 💡 Computers have been traditionally bad at understanding chains of intents
• Utterance: a phrase or way in which an intent might be said
  • Intents are trained with sample utterances
  • e.g. “Can I order a hamburger?”, “I want a burger”, “Cheesburger, please”…
• Slot: key parameters inside an utterance
  • e.g. hamburger size (small/medium/large), meat (pork/beef)