Virtual Conversations: An Interface to Knowledge

effort revolutionized spoken language technology and made independent speaker and continuous, real-time speech recognition systems a reality on large, powerful systems. Subsequent advancements in speech recognition technology by IBM, Dragon Systems, Learnout and Hauspie, Entropics, and other companies, resulted in the capabilities of independent speaker and continuous speech, real-time recognition in a PC environment. Researchers have made numerous accomplishments in natural language dialogue models since the speech recognition breakthrough. Most, if not all, of these important developments involve avatars. These approaches include the Virtual Humans and Conversational Agents process embodied in the Research Triangle Institute's Avatalk 1 and the Center for Spoken Language Understanding (CSLU) Speech Toolkit. 2 In these models, the computer side of the conversation is constructed for task-specific purposes. Much work has also been done on the social and psychological implications of human–com-puter interactions 3-5 (see also In this article, we describe the Interactive Drama Inc. (IDI) voice-activated multimedia model called Conver-sim (conversation simulator), which lets a user conduct a virtual interview with a real person whose video images have been digitized and stored on a PC (see series of Delphi programs that incorporate state-of-the-art speech recognition and digital video technologies and operate on a standard PC. The user speaks naturally into a microphone attached to the computer to engage the video character in a face-to-face, simulated conversation. For more information about Conversim, see the " Technical Note " sidebar. A Conversim dialogue typically begins when the user picks up a microphone and says, " Begin the session. " The featured character comes to life onscreen and says " Hello. " If necessary, the system displays a brief paragraph of text describing who the person is and the general subject of the conversation. A simple instruction tells the user to speak one of the questions scrolling at the bottom of the screen. ConverScroll, IDI's intelligent, nondirective prompting system displays these questions. The user can ask any of these questions, using the wording provided, or paraphrasing it, to activate the conversation. Instantaneously, the video character answers the question and the dialogue begins. This beginning is all that's necessary to orient users. Users intuitively conduct the conversation from this point, taking advantage of the scrolling prompts. The user's voice controls the entire experience; the user never has to touch a mouse or a keyboard. Figure 1 illustrates one possible interface. It's from a Conversim program about biological warfare featuring …