We investigate the effectiveness with which the accuracy of a prompted speech corpus can be validated when minimal additional speech resources are available, and specifically when a language model in the target language ...