Emotion recognition from helpdesk messages

This paper describes system for emotion recognition which can be used to determine the priority of messages on the first level of helpdesk services. An algorithm used in this paper uses artificial intelligence (SVM classifier) and can recognize 5 different emotions. The used emotional classes were based on acoustic model which was inspired by acoustic emotion recognition research works. The proposed system has evaluated 5 classifiers and identifies a dominant emotion class. This work also describes a small database which was created on the basis of the selected helpdesk messages. The database was used in training and testing of the mentioned classifier. Success of classifier achieved in this work is 76.63% and impact of the proposed optimization methods on the final model accuracy has been proven.