Decoder. plain-English AI glossary

ASR

● Core

Automatic speech recognition—converting audio directly to text without human transcription.

Think of it like

A secretary who listens and types perfectly, understanding accents and background noise.

Example

Google Assistant converts "Play jazz music" from your voice directly to text command.

How it actually works

Can use hybrid models (acoustic model for audio features + language model for text) or end-to-end (CTC or attention seq2seq). Language model helps correct speech recognition errors via context. Robust to accents, noise, and speaking rate if trained on diverse data.

For product teams

Core capability for voice interfaces and accessibility (closed captions).

For engineers

Acoustic features extracted from audio; HMM or neural net aligns to phones; language model re-ranks hypotheses.

Speech-to-Text — Synonym for.

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open Decoder Add to Chrome

ASR