Novel Approaches to Pattern-based Interaction Quality Modeling
In the previous chapters it has been shown that users show emotions towards SDS. However, it has also become obvious that only a minority of the users reacts emotionally while interacting whereas the vast majority often remains emotionless. This user behavior seems natural as users do not expect systems to be able to recognize emotions. While the presence of emotions may allow inferences to be drawn on user satisfaction, the absence of emotions consequently may not. A user may be dissatisfied without showing anger and satisfied without showing happiness, i.e., an interaction may be endangered without the user showing emotions. As a result, other (complementary) approaches to emotion recognition are required for online monitoring SDS in order to detect critical dialog situations.