Dear all,

LIUM (Laboratory of Informatics of the University of Le Mans, France) distributes under Creative Commons BY-NC-ND 3.0 license a corpus based on TED talks (http://www.ted.com) and especially designed to estimate acoustic models in English.

It contains:

- about 118h of speech 

- 799 audio talks in NIST sphere format (SPH) 

- 799 transcripts in STM format 

- Dictionary with pronunciation (157617 words) 

More details are here:

http://www-lium.univ-lemans.fr/fr/content/corpus

Best,

Yannick

Hi Yannick,

thanks for letting us know,

Ken

Link for TED-LIUM has been updated, here is the new address:

http://www-lium.univ-lemans.fr/fr/content/corpus-ted-lium

Best,

Anthony