The VoxForge corpus was used in research by Gregoire Montavon, Machine Learning Group, Berlin Institute of Technology, Germany on  Deep learning for spoken language identification.  From the abtract:

Empirical results have shown that many spoken language identification systems based on hand-coded features perform poorly on small speech samples where a human would be successful.  A hypothesis for this low performance is that the set of extracted features is insufficient. A deep architecture that learns features automatically is implemented and evaluated on several datasets.

--- (Edited on 12/15/2009 5:52 pm [GMT-0500] by kmaclean) ---