liveBook · Manning

Speech recognition

As explained in Architecting Accessibility

Automated speech recognition is technology that converts spoken language into text without requiring a human transcriber to produce each word.

Automated speech recognition (ASR) uses an AI engine to transcribe speech in real time. For live events, it can provide captions or transcripts without a human transcribing every word. section 3.4.2

Choose an approach with both access and accuracy in mind:

  • ASR is more efficient than human-driven captioning, but transcript quality may require monitoring. section 3.4.2
  • Hybrid captioning combines AI-generated captions with real-time human review and correction. section 3.4.2
  • Whichever approach is used, captions or transcripts need to be programmatically accessible to assistive technologies. section 3.4.2

How Architecting Accessibility covers it

How deep it goes
Explains real-time AI transcription and its efficiency–accuracy trade-off, but does not detail how to build an ASR engine.
Examples
Uses live video streams and live audio-only events as settings for real-time speech-to-text.
What you can do after
Evaluate whether ASR’s efficiency suits a live event and whether monitoring or human review is needed.
How it connects
Connects speech recognition with live captions, transcripts, and assistive technologies.

Also explained in

Compare how 8 books explain speech recognition →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Architecting Accessibility. The section links point to the parts of the book it draws on.