This is the TED-LIUM corpus release 3, licensed under Creative Commons BY-NC-ND 3.0.

All talks and text are property of TED Conferences LLC.

This new TED-LIUM release was made through a collaboration between the Ubiqus company and the LIUM (University of Le Mans, France)

Contents:

• 2351 audio talks in NIST sphere format (SPH), including talks from TED-LIUM 2: be careful, same talks but not same audio files (only these audio file must be used with the TED-LIUM 3 STM files)
• 452 hours of audio
• 2351 aligned automatic transcripts in STM format
• TEDLIUM 2 dev and test data: 19 TED talks in SPH format with corresponding manual transcriptions (cf. ‘legacy’ distribution below).
• Dictionary with pronunciations (159848 entries), same file as the one included in TED-LIUM 2
• Selected monolingual data for language modeling from WMT12 publicly available corpora: these files come from the TED-LIUM 2 release, but have been modified to get a tokenization more relevant for English language

Two corpus distributions:

• the legacy one, on which the dev and test datasets are the same as in TED-LIUM 2 (and TED-LIUM 1).
• the ‘speaker adaptation’ one, especially designed for experiments on speaker adaptation.

More details are given in this paper:

François Hernandez, Vincent Nguyen, Sahar Ghannay, Natalia Tomashenko, and Yannick Estève, “TED-LIUM 3: twice as much data and corpus repartition for experiments on speaker adaptation”, submitted to the 20th International Conference on Speech and Computer (SPECOM 2018), September 2018, Leipzig, Germany

A preprint version is available on arxiv (and in the doc/ directory).

SPH format info:
Channels: 1
Sample Rate: 16000
Precision: 16-bit
Bit Rate: 256k
Sample Encoding: 16-bit Signed Integer PCM