NOVEMBER 17, 2026

BEL Female Talker Dataset

Audio and audio-visual recordings of the 500 Basic English Lexicon (BEL) sentences produced by one adult female talker at average and high vocal effort (Lombard speech), arranged in 20 reshuffled lists of 25 sentences.

Frame from a BEL Female Talker audio-visual recording

Overview

The BEL Female Talker dataset contains audio and audio-visual recordings of the 500 Basic English Lexicon (BEL) sentences (Calandruccio & Smiljanic, 2012), produced by one adult female native speaker of American English in two speaking styles: average vocal effort and high vocal effort (Lombard speech). Each of the 1,000 recordings is provided as a 48 kHz, 16-bit WAV file and as a frontal-view MP4 video (3840 x 2160, 60 fps) whose audio track is identical to the WAV file.

The sentences are arranged in 20 reshuffled lists of 25 sentences (BELR lists), with four scored keywords per sentence. The download also includes speech-shaped noise matched to each speaking style, a table of list assignments and scoring keywords, and the analysis code used to build the lists and derive critical differences. The list assignments and code are also on GitHub.

The audio recordings and BELR lists were validated in a speech recognition study with normal-hearing adult listeners (Gijbels, Johnson & Miller, in review). The audio-visual recordings have not yet been validated.

For questions or feedback, please reach out to lgijbels@meta.com.

License

The recordings, noise files, list assignments, and analysis code are released by Meta Platforms, Inc. under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0). The BEL sentence texts are from Calandruccio & Smiljanic (2012).

Citation

If you use this dataset, please cite:

Gijbels, L., Johnson, K., & Miller, C. (2026). Validation, critical differences, and list optimization of Basic English Lexicon (BEL) sentences across talkers and vocal effort levels. Journal of Speech, Language, and Hearing Research. [volume, pages, and DOI to be added on publication]

Calandruccio, L., & Smiljanic, R. (2012). New sentence recognition materials developed using a basic non-native English lexicon. Journal of Speech, Language, and Hearing Research, 55(5), 1342–1355. https://doi.org/10.1044/1092-4388(2012/11-0260)