Conversational Agents

Understanding conversational utterances often requires techniques that are very different from those used in understanding well formed text. Conversations usually consist of short utterances that are often incomplete and grammatically incorrect. Understanding these utterances depends heavily on knowing the context of the conversation. For these reasons, pragmatics, or world knowledge is extremely important. In practice, when a large amount of pragmatic knowledge is available, syntactic and semantic analysis become less important. The approach taken in most current, as well as past, implementations of conversational agents, is to avoid detailed linguistic analysis of utterances and instead use relatively simple pattern matching, based on domain knowledge. However, many domain-specific patterns must be written to provide credible interactions with the user. Since domain knowledge is essential for a conversational agent, the approach taken here is to use pattern matching as the simplest and most direct approach.

This work is related to another project on intelligent environments.

Selected Publications

Sammut, C. (2001). Managing Context in a Conversational Agent. Linkoping Electronic Articles in Computer and Information Science. 6(27).

Mak, P., Kang, B.-H., Sammut, C. and Kadous, M. W. (2004). Knowledge Acquisition Module for Conversational Agents. In C. Zhang, H. W. Guesgen and Wai K.Yeap (Eds.), PRICAI 2004: Trends in Artificial Intelligence (Vol. 3157): Springer.

Mohammed Waleed Kadous and Sammut, C. (2004). InCA: A Mobile Conversational Agent. In C. Zhang, H. W. Guesgen and W. K. Yeap (Eds.), 8th Pacific Rim International Conference on Artificial Intelligence (pp. 644 - 653). Auckland, New Zealand: Springer.