MITLL 2015 Language Recognition Evaluation System Description
MIT Lincoln Laboratory Lexington United States
Pagination or Media Count:
The training data segments for the mandatory fixed training task made available by NIST was derived primarily from previously released data sources Callhome, Callfriend, Mixer3, Switchboard, and VOA. Error Reference source not found. shows a breakdown of the numbers of cuts and the speech duration post-SAD available for each language in the fixed training data set.