Assessing the Impact of Adaptive Generation in the COMIC Multimodal Dialogue System
Mary Ellen Foster · 2005
We describe how information from the dialogue history and the user model is incorporated into the output-planning process of the COMIC multimodal dialogue system, and present the results of experiments analysing the impact of both of these factors on the generated descriptions. The results of the experiments confirm and extend previous results by showing that both forms of adaptation make a perceptible difference in the generated speech. They also point to ways in which the system output can be improved to take further advantage of these information sources; in particular, they indicate the importance of concessions to negative preferences to the perceptibility of user-model tailoring. 1