When, for example, a call is made to the telephone company, the call is taken by just such a system. The caller is connected to the appropriate assistant according to the reason for the call. One is greeted by such a system even if one calls IKEA. Among other things, it allows one to obtain information on product availability.
But what exactly is such a speech portal? How does it work, and how is it set up? With a speech portal, callers are able to carry on a natural language dialogue by partially or fully automated means through a telephone or other acoustic medium.
The dialogue and all of its defined variants are described in the VoiceXML. This dialogue description is processed during the runtime by the dialogue flow interpreter. The system’s speech utterances are either pre-recorded or generated by the speech dialogue system’s speech synthesiser (text-to-speech engine, TTS). The caller’s utterances are understood by the speech dialogue system’s speech recognising system (Automatic Speech Recognition, ASR). Information needed during a dialogue with the caller is read by the back-end systems. Information that is received, such as purchase order information, is transmitted to the back-end systems.
Speech dialogue systems are especially good at mapping simple dialogues. These include, for example:
Speech dialogue systems can be called up (inbound), but they are also able to make calls (outbound) to, for example, remind retailers of a weekly order or to call out the fire brigade.
Speech dialogue systems are especially well suited for situations in which computers are not available or cannot be used. They are:
For the user to accept the system, however, a dialogue design that matches the application is indispensable.