Skip to content

LibriSpeech what it measures, how it is scored and who leads it

Speech · Johns Hopkins University · introduced 2015 · Word error rate · saturated

About this benchmark

Task
Transcribe read English audiobook speech.
Dataset
About 1,000 hours of LibriVox audiobooks; clean and other test sets.
Method
Word error rate against the reference text.
Metric
Word error rate
Organization
Johns Hopkins University
Introduced
2015

Versions

Newest first. New versions are added, never rewritten.

VersionDate
1Published at ICASSP 2015.201511 years ago

Sources: each benchmark's paper (arXiv) and its official site or leaderboard. Scores are the boards captured here every day, exactly as published.

New #1s by email

Saturdays, only in weeks when a leaderboard has a new #1.

Double opt-in. Unsubscribe any time.