AWS speech AI services convert audio to text and text to audio, while conversational AI services enable building chatbots and intelligent contact centres. Amazon Connect integrates these capabilities for enterprise-grade customer service AI.

Key Points

  • Amazon Transcribe: speech-to-text; supports real-time streaming and batch; speaker diarisation, custom vocabulary
  • Transcribe Medical: clinical audio transcription — medical terms, physician-patient conversations
  • Transcribe Call Analytics: analyse call centre recordings — sentiment, talk time, interruptions, action items
  • Amazon Polly: text-to-speech with 60+ voices in 30+ languages; SSML for pronunciation control; Neural TTS
  • Amazon Lex: conversational AI (NLU) — intents, slots, fulfilment with Lambda; powers Amazon Alexa
  • Lex Automated Chatbot Designer: generate Lex bot from conversation transcripts automatically
  • Amazon Connect: cloud contact centre; integrates Lex (bots), Transcribe (recording), Comprehend (sentiment)
  • Amazon Q in Connect: real-time agent assistance — surfaces relevant knowledge during live calls
  • Amazon Chime SDK: add voice/video to applications; includes noise suppression and voice focus ML
ServiceDirectionKey FeatureExam Trap
TranscribeAudio → TextSpeaker diarisation, custom vocabularyNot Polly — Polly is text→audio
PollyText → AudioNeural TTS, SSML, 60+ voicesCannot transcribe — only synthesise
LexText/Voice → IntentSlots, fulfilment, multi-turnLex uses NLU; Kendra uses NLP search
ConnectContact centreIntegrates all speech servicesConnect orchestrates; services do the work

Real-World Example

Capital One uses Amazon Lex in their voice IVR to handle millions of calls per month — customers can check balances, make payments, and dispute charges by voice, without reaching a human agent. Transcribe Call Analytics reviews 100% of agent calls for quality assurance.