15.1 Development

The development phase includes the creation of the application software, the development of the grammars to be used by the recognizer, and the production of the audio for the prompts and any nonverbal audio that is part of the application.

15.1.1 Application Development

Application software development entails the coding of the dialogs as well as integration with backend databases and other software systems.

Application logic that implements the dialog can be written in a number of languages. Occasionally it is written in standard programming languages such as C, C++, or Java. More commonly it is coded in a proprietary language associated with an interactive voice response (IVR) platform that is designed to run touchtone and speech applications over the telephone network. Many IVR platforms have their own proprietary tools and languages that developers must use to implement applications.

A growing trend, which is likely to become the dominant implementation approach in the not-too-distant future, is to code the dialog in VoiceXML, a markup language for building speech interfaces. It is similar in spirit to HTML, the markup language often used for creating Web sites.

One reason for interest in VoiceXML is that it is becoming a widely used standard controlled by the World Wide Web Consortium (W3C). Furthermore, it is a speech-centric language, with basic elements such as menus (presentations to the caller of a set of choices) and forms (used to collect information from the caller, often handling multiple pieces of information in a single form). VoiceXML is an XML-based language and thus leverages the Web development paradigm already familiar to many programmers. There are a number of excellent references on VoiceXML, including Larson (2003), Sharma and Kunins (2002), and Abbott (2002).

In addition to the dialog code, application development involves the integration of your application with other software systems such as backend databases, computer telephony integration (CTI) software, and Web services. In some cases for example, if the speech system is replacing a touchtone system the integration code may already exist. Most IVR platforms support integration with databases and CTI systems.

15.1.2 Grammar Development

The development of effective grammars those that have high coverage of things callers are likely to say to the system is key to the development of a usable system. Grammar development is often a task of the VUI designer. Even if someone else develops the grammar, there is an intimate relationship between the grammar, the prompt wording, and the chosen dialog strategies. Chapter 16 covers the development and tuning of grammars in detail.

15.1.3 Audio Production

Audio production involves the recording of all the prompts by an appropriate voice actor. The designer often acts as the voice coach at the recording sessions, working with the voice actor to create the persona. The coach must also make sure that the voice actor understands how all the prompts function in context. Prompts that are to be concatenated must be recorded with appropriate prosody so that concatenation splices can be seamless. Chapter 17 covers the details of preparation and coaching of recording sessions with voice actors.

If the application uses nonverbal audio, final recordings must be prepared and processed. Final recordings should be tested over the phone.