Emotion analysis in DiaCoSk dialog corpus

Abstract

The proposed paper describes our experimental work with the dialogue corpus DiaCoSk, which is a small corpus of dialogue interactions in the Slovak language. The original corpus consists of recordings, transcriptions and dialogue act labels that can be used for training dialogue act labeling machine. Emotions in DiaCoSK utterances were labeled additionally to observe their frequency and relation to dialogue acts. Obtained results indicate interesting relations between dialogue acts and emotions. Moreover, influence of emotions on automatic DA labeling has been proven.