qiujiali/lattice_rnn

Bi-directional Lattice Recurrent Neural Networks for Confidence Estimation

37
/ 100
Emerging

This project helps speech recognition engineers improve the accuracy of confidence scores from automatic speech recognition (ASR) systems. It takes speech 'lattices' or 'confusion networks' from your ASR output and provides more reliable confidence predictions, not just for the most likely word sequence but for all alternative word hypotheses. This is ideal for ASR system developers, researchers, or anyone building applications that rely on highly accurate confidence estimation for spoken language.

No commits in the last 6 months.

Use this if you need to significantly improve the confidence scores generated by your automatic speech recognition (ASR) system, especially when working with alternative word hypotheses beyond just the single-best prediction.

Not ideal if you are looking for an off-the-shelf, end-to-end ASR system or a solution that doesn't require pre-processed speech 'lattices' or 'confusion networks' as input.

speech-recognition ASR-confidence-scoring spoken-language-processing natural-language-understanding
Stale 6m No Package No Dependents
Maintenance 0 / 25
Adoption 6 / 25
Maturity 16 / 25
Community 15 / 25

How are scores calculated?

Stars

15

Forks

5

Language

Python

License

MIT

Last pushed

Aug 28, 2020

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/voice-ai/qiujiali/lattice_rnn"

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