Speech Recognition

Speech recognition, also called automatic speech recognition (ASR) or speech-to-text, converts spoken audio into text. Current systems typically use deep neural networks that map audio features directly to characters or tokens, often combined with a language model that favours likely word sequences.

Why it matters

It powers voice assistants, captions, meeting transcripts, dictation and call-centre analytics. Accuracy is commonly reported as word error rate (WER): substitutions, deletions and insertions divided by the number of reference words. Performance drops with background noise, strong or unfamiliar accents, overlapping speakers, domain jargon and poor microphones, so results should be tested on audio resembling your real use. Recording people also raises consent and privacy questions that vary by country.

Example

Reference:  "restart the web server now"
Recognised: "restart the web survey now"
Errors: 1 substitution in 5 words -> WER = 20%

Common mistakes

• Judging a system on clean studio audio only.
• Ignoring speaker separation (diarisation) when several people talk.
• Storing transcripts containing personal data without a retention policy.

Related

• Text-to-speech
• NLP
• Deep learning
• PII

This page is general information written by our editorial team and updated when documentation changes. Servers and providers differ, so test changes safely first. See our editorial policy and disclaimer.