Although there are many different capability profiles of automatic speech recognition systems, we will now concentrate on two examples that span a wide range of systems, namely the isolated word recogniser and the large vocabulary continuous speech recogniser in the following section.
We will implicitly assume that the isolated word recogniser has some other feature that makes it attractive for an application: ``real-time'' processing of the speech input. This does not necessarily mean that the recognition process happens in 1:1 real time, but we merely want to exploit the fact that isolated word recognition systems can give one word of output on each input word, so that the assessment process can be interactive.