DanielLin94144/Test-time-adaptation-ASR-SUTA

Test-time adaptation for speech recognition model by single utterance. The official implementation of "Listen, Adapt, Better WER: Source-free Single-utterance Test-time Adaptation for Automatic Speech Recognition" paper.

30
/ 100
Emerging

This project improves the accuracy of speech recognition models when they encounter unfamiliar accents, background noise, or speaking styles. It takes an audio recording and an existing speech-to-text model, then adapts the model in real-time to transcribe that specific recording more accurately. This is designed for researchers and engineers working with automatic speech recognition (ASR) systems.

No commits in the last 6 months.

Use this if you need to improve the word error rate of an existing ASR model on individual, diverse audio inputs without retraining the entire model or needing the original training data.

Not ideal if you are looking for a pre-built, production-ready speech-to-text service or if your ASR model is not based on the CTC architecture.

speech-to-text audio-transcription speech-recognition-research model-adaptation
No License Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 6 / 25
Maturity 8 / 25
Community 16 / 25

How are scores calculated?

Stars

20

Forks

6

Language

Python

License

Last pushed

Apr 01, 2022

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/voice-ai/DanielLin94144/Test-time-adaptation-ASR-SUTA"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.