Speech: captions on a video call
How speech recognition changes sound into words, what helps it work and why it can mishear people.
Speech recognition case study
What this example shows
Follow spoken audio through a speech-recognition system and understand why caption errors occur.
What goes in
Audio captured by the call, including speech, background sounds, overlapping voices and the quality of each person’s microphone.
What the AI system does
A speech-recognition model identifies patterns in the sound and predicts words. The service may use language patterns to choose between words that sound alike.
What comes out
Time-aligned written captions displayed during or after the conversation.
What can go wrong
Noise, unfamiliar names, accents, poor connections and several people speaking can increase errors. Incorrect captions can change the meaning of important information.
Who is responsible
The service should make caption limitations clear and support corrections. Participants should confirm important names, numbers and decisions through another route.