Natural Human-System Interaction Using Intelligent Conversational Agents

Background

Spoken dialogue constitutes the most natural and powerful means to interact with computers. Systems based on natural spoken dialogue start to appear feasible with the recent improvements in computer engineering and in speech and language processing. Speech-based interfaces increasingly penetrate into environments that can benefit from hands-free and/or eyes-free operation. In the context of the growth of network-based information services, messaging and edutainment, or the demand for personalized real-time services, automatic speech recognition and speech synthesis are highly promising. They are considered as sufficiently mature technologies to allow their inclusion as effective modalities in both telephony and multimodal Human-Computer Interaction (HCI) (Deng & Huang, 2004). Some applications of Internet searching and navigating are currently known; Opera version 9 has a basic voice interface (Opera, 2007). However, its recognition engine is pretty basic and cannot be trained; also, its synthesized voice is robotized. Google provides this kind of functionality through Google Voice Search which is still in a demo state. To use it, a user needs to make a phone call. This is not very convenient for users (Google, 2006). Another application is the one developed by Dr. Meirav Taieb-Maimon and colleagues from Ben-Gurion University of the Negev where car drivers can consult the Internet through voice commands (Sommer, 2005). Lyons et al. (2004) introduced a concept of a dual-purpose speech interaction that provides meaningful input to the computer in order to manage calendar and other communication tools for users. In these applications, the limitation of automatic speech recognition engines is outlined as the main obstacle to the efficiency of the interaction.