Research on Sentiment Classification for Tang Poetry based on TF-IDF and FP-Growth

Ancient poetry is an excellent traditional culture through chronic historical accumulation in China. One of the most critical issues in studying poetries is how to effectively understand the quintessence and emotions contained in them. This paper presents a novel emotion automatic annotation strategy particularly answer this question. First, we use the Valence-Arousal space model to realize the continuity of the ancient poetry's emotional expression based on traditional classifications of ancient poetries. Second, we propose an emotional characteristics extracting algorithm which makes both use of TF-IDF's good support for feature extraction and FP-growth algorithm's speciality in mining relationships between words. Finally, back propagation neural network is used as a classifier to automatic label the ancient poetry emotion. The correctness, stability and practicability of the proposed strategy are proved with both theoretical analysis and experimental simulations. The evaluation results show that the proposed strategy can achieve automatic annotation of emotions under acceptable accuracy.