Download zip Select Archive Format
Name Last Update history
File empty ..
File dir s5 Loading commit data...
File txt README.txt Loading commit data...


This is a Kaldi recipe for HUB4 Spanish Broadcast News (HUB4-NE).
Uses the corpora LDC98T29 (transcript), LDC98S74 (speech) and 
LDC2001S91 (eval data)

This recipe uses a graphemic lexicon generated directly from the transcripts
(i.e. no other sources of phonetic knowledge is needed). 
The amount of tranining audio is approximately 30 hours.