Segment-based Fine-grained Emotion Detection for Chinese Text

Emotion detection has been extensively studied in recent years. Current baseline methods often use token-based features which cannot properly capture more complex linguistic phenomena and emotional composition in fine grained emotion detection. A novel supervised learning approach―segment-based fine-grained emotion detection model for Chinese text has been proposed in this paper. Different from most existing methods, the proposed model applies the hierarchical structure of sentence (e.g., dependency relationship) and exploits segment-based features. Furthermore, the emotional composition in short text is addressed by using the log linear model. We perform emotion detection on our dataset: news contents, fairly tales, and blog dataset, and compare our proposed method to representative existing approaches. The experimental results demonstrate the effectiveness of the proposed segment-based model.