It is very likely that a large vocabulary recognition system has not yet been fully tuned for the application. In order to make it possible for the system to be tuned, some development test material is required. In principle this material should be very similar to the assessment test material, but of course essential details should differ. The utterances, for instance, should not be the same, and if the system is speaker independent, there should be no overlap in speakers. If no special recordings have been made for development testing, one could split off a fraction of the training material and use that for development testing.