Interaction Profiles for an Artificial Conversational Companion

Using Artificial Companions for tasks requiring long-term interaction like language learning or coaching can be approached by creating local computational models for particular interaction structures, and models reflecting changes in interaction over time. An Artificial Conversational Companion (ACC) that helps to practice conversation in a foreign language is expected to play the role of a language expert in conversation. We apply methods of Conversation Analysis to obtain datadriven models of interaction profiles for language experts and language novices from a corpus of instant messaging based dialogues between native speakers of German and advanced learners of German as a foreign language. We show different ways how the artificial agent can simulate ”doing being expert” in conversation and promote learning.