AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Scaling Speech Technology to 1,000+ Languages

Vineel Pratap, Andros Tjandra, Bowen Shi, Paden Tomasello, Arun Babu, Sayani Kundu, Ali Elkahky, Zhaoheng Ni, Apoorv Vyas, Maryam Fazel-Zarandi, Alexei Baevski, Yossi Adi and 4 more

JMLR 2024 · Large language models › Language, safety and interpretability · approach: Sign language datasets and translation

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
pre-train on 1,406 languages (MMS) instead of 128 languages (XLS-R), 300M sizewav2vec 2.0 ASR · FLEURS-61CER: 0.6 lower CERImproved
pre-train on 1,406 languages (MMS) instead of 128 languages (XLS-R), 1B sizewav2vec 2.0 ASR · FLEURS-61CER: 0.7 CER differenceImproved
scale parameters from 300M to 1BXLS-R · FLEURS-61CER: 3.2 CER decreaseImproved
scale parameters from 300M to 1BMMS-lab · FLEURS-61CER: 3.0 CER improvementImproved
scale dense ASR from 61 to 1,107 languagesdense MMS ASR · FLEURS-61CER: +5.1 CERHurt
scale dense ASR from 61 to 1,107 languagesdense MMS ASR · CV-49CER: +2.1 CERHurt
add language-specific adapters and heads (LSAH) (ablation)MMS ASR scaled to 1,107 languages · FLEURS-61CER degradation: 5.1 -> 0.4Improved
add language-specific adapters and heads (LSAH) (ablation)MMS ASR scaled to 1,107 languages · CV-49CER degradation: 2.1 -> 0.2Improved
use MMS-lab CTC model with n-gram LM instead of WhisperMMS ASR · FLEURS-54WER: 58% relative WER reductionImproved
scale LID from 126 to 4,017 languagesMMS LID · BABELaccuracy: -3.6%Hurt
scale LID from 126 to 4,017 languagesMMS LID · FLEURSaccuracy: -0.3%No clear effect
train LID on MMS-lab-U+unlab instead of in-domain VoxLingua-107MMS LID · FLEURS (72 languages)accuracy: -2.1%Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy