Advanced research in applied artificial intelligence, Lecture Notes in Computer Science
Abstract:Automatic emotion sensing in textual data is crucial for the development of intelligent interfaces in interactive computer applications. This paper reports a high-precision, domain-independent approach for automatic emotion sensing for “events” embedded in sentences. The proposed approach is based on the common action distribution between the subject and object of an event. We have incorporated semantic labeling and web-based text mining techniques, together with a number of reference entity pairs and hand-crafted emotion generation rules to realize an event emotion detection system. Moreover, a hybrid emotion detection engine is presented by incorporating a set of predefined emotion keywords and the proposed event-level emotion detection engine. The evaluation outcome reveals a rather satisfactory result with about 73% accuracy for detecting the Happy, Sad, Fear, Angry, Surprise, Disgust, and Neutral.