XbotML: A markup language for human computer interaction via chatterbots

Abstract

Since 1995, we have been watching the emergence of a new paradigm for the construction of chatterbots based on markup languages. The most prominent of these languages is AIML. Despite its success, XML chatterbots have drawbacks in what concerns the level of fluency in dialogues. We present here XbotML, a new language for the construction of chatterbots based on principles of the Conversational Analysis theory. Following this theory, XbotML structures dialogues between user and chatterbot in adjacency pairs, each pair bearing one associated intention. As proof of concept, we constructed a chatterbot that has reached fluency level around 75% in dialogues with people. This is an original work that contributes to raise the fluency level of XML chatterbots by: providing a linguistically grounded model for chatterbots' markup languages bases; providing a modular model for these bases; and allowing the extension of existing bases to different domains and applications. © Springer-Verlag Berlin Heidelberg 2003.