Speech-recognition applications are comprised of two fundamental components: the speech recognizer and the VUI. The speech-recognition engine is responsible for converting audible sounds into words or phrases. The VUI applies meaning to the words converted by the speech recognizer and guides users through an interactive experience. These two components work in concert in an attempt to create a pleasant user experience. For voice-portal subscribers seeking access to news, information, and personal communications, the most fundamental element of the user experience is the VUI because it is vital to enabling easy navigation.

Menu-Based VUIs

Traditionally speech-enabled applications were menu-based or hierarchical in nature. A menu-based VUI accepts a list of limited responses depending on where the user is in the system. Similar to traditional automated attendants found at major corporations, menu-based VUIs offer a step-by-step process to accomplish basic functions. These VUIs were the result of a technology transition from touch-tone to voice-recognition interfaces. This transition resulted in a limited interface design because communication by touch-tone is much more systematic and methodical than the way in which people naturally interact.

For example:

Users have expressed many concerns with menu-based VUIs, including having to negotiate excessive steps, getting stuck in submenus, losing track of where they are in the process, and speaking in unnatural, short spurts. Similar to a maze, it is very easy to take a wrong turn with a menu-based VUI and to not know where to go next. In addition, nonintuitive key words such as phone number must be memorized to activate certain functions. In natural speech, users would never say, "phone number" and expect someone to understand that the phrase really means, "Dial a phone number" or "Dial 555-1212."

Moreover, menu-based VUIs make it very difficult to change functions once a user has activated a menu substructure. For example, if the system requested a phone number and the user decided to listen to his or her remaining voice messages, the system would not know how to respond to a "play messages" command because it was expecting phone numbers. This can lead to a continuous loop and extreme user frustration.

There is no doubt that menu-based VUIs are beneficial in highly controlled environments that have a limited number of possible commands, such as an alternative to touch-tone–based customer-service centers. However, more complex applications, such as voice portals, require a conversational VUI.