CMU Wilderness Multilingual Speech Dataset

By Alan W. Black. 2019

115 citations recorded by the source corpus.

This paper describes the CMU Wilderness Multilingual Speech Dataset. A dataset of over 700 different languages providing audio, aligned text and word pronunciations. On average each language provides around 20 hours of sentence-lengthed transcriptions. We describe our multi-pass alignment techniques and evaluate the results by building speech synthesizers on the aligned data. Most of the resulting synthesizers are good enough for deployment and use. The tools to do this work are released as open source, and instructions on how to apply such alignment for novel languages are given.

Subfields: Artificial Intelligence, Evaluation & Benchmarks, Speech & Audio, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Evaluation & Benchmarks

In Speech & Audio

In Signal Processing

Citations in the corpus

FLEURS: FEW-Shot Learning Evaluation of Universal Representations of Speech (2022, background): “6k hrs YouTube Spontaneous No No No CMU Wilderness [23] 700 14k hrs Religion Read Yes Yes Yes FLEURS (this work) 102 1.”
VoxPopuli: A Large-Scale Multilingual Speech Corpus for Representation Learning, Semi-Supervised Learning and Interpretation (2021, background): “The CMU Wilderness dataset (Black, 2019) collects readings from the New Testament, with 700 different languages available.”
MLS: A Large-Scale Multilingual Dataset for Speech Research (2020, methodology): “Other notatable multi-lingual datasets distributed under an open license are the M-AILABS [5] and the CMU Wilderness [6] datasets.”
Libri-Light: A Benchmark for ASR with Limited or No Supervision (2019, neutral): “Black at CMU has compiled the Wilderness dataset which consists of the text of the Bible read in 750 languages [13].”
Massively Multilingual Adversarial Speech Recognition (2019, methodology): “In this paper we explore pretraining multilingual ASR models using speech from as many as 1 voxforge.org 100 languages from the CMU Wilderness Multilingual Speech Dataset (Black, 2019).”
Open-source Multi-speaker Speech Corpora for Building Gujarati, Kannada, Malayalam, Marathi, Tamil and Telugu Speech Synthesis Systems (2020, methodology): “CMU Wilderness Dataset This speech dataset consists of aligned pronunciations and audio for about 700 different languages based on readings of the New Testament by volunteers (Black, 2019).”
XTREME-S: Evaluating Cross-lingual Speech Representations (2022, methodology): “LangID can be evaluated using VoxLingua107 [25] on YouTube data in 107 languages, and CMU Wilderness [26] on New Testament data in 700+ languages.”
Scaling Speech Technology to 1, 000+ Languages (2023, methodology): “The most related project is the CMU Wilderness effort [Black, 2019] which created speech synthesis models for 699 languages using speech and text data from similar sources as our datasets.”
MaSS: A Large and Clean Multilingual Corpus of Sentence-aligned Spoken Utterances Extracted from the Bible (2019, background): “The CMU Wilderness corpus [1] is a speech dataset containing over 700 different languages for which it provides audio excerpts aligned with their transcription.”