Comparing the performance of knowledge-based and machine-learning approaches for the detection of emotions in an english Text

Abstract

The detection of emotion and sentiment analysis are very hot topics at the moment and the detection of emotion from written text still remains a difficult subject of this area of research. The main approaches to this task are knowledge-based approaches and machine-learning approaches. This paper examines the performance of two approaches (a knowledge-based and a machine-learning approach) on a small corpus of chat text annotated with emotion labels. It will be shown that the machine-learning approach used in this experiment outperforms the knowledge-based approach in all aspects.