Speech Recognition
Transcribing spoken audio into text.
Overview
Speech recognition converts audio into text, and modern systems do it end to end from waveform to characters rather than by chaining separate acoustic and language models. It solves the interface problem of hands-free and eyes-free use. It exists as one of the earliest successful neural applications, and its accuracy now depends mostly on audio quality, domain vocabulary and latency budget rather than on the modelling approach.
Technology Graph
How Speech Recognition connects to the rest of the stack.
Uses and depends on
- Attention concept Algorithms Speech Recognition uses Attention High ✓ source
- Transformer architecture Algorithms Speech Recognition based on Transformer High ✓ source
Applied to
- Whisper model Multimodal Whisper used for Speech Recognition High ✓ source
- llama.cpp tool Infrastructure llama.cpp used for Speech Recognition Medium ✓ source