面向高校学生微博的跨粒度情感分析

传统的微博情感分析往往忽略不带感情色彩的情感词对微博情感的影响，并缺乏对复杂句式的分析。为此，提出结合条件随机场（conditional random filed，CRF）和复杂句式的跨粒度情感分析方法。该方法在CRF模型的基础上，融合复杂句式特征和语义依存特征，对学生微博进行细粒度情感分析，识别出微博文本中的情感要素，在此基础上，通过基于复杂句式的粗粒度情感分析方法分析微博文本的情感倾向，实现对学生总体情感倾向的跨粒度分析。实验结果显示，跨粒度情感分析方法的提出，使得情感要素识别的综合准确率达88%左右，微博情感分析的综合准确率达87%左右。比起传统的情感分析方法，准确率更高，分类效果更好。

Cross-grained sentiment analysis oriented to college student micro-blog

Traditional sentiment analysis of micro-blog often ignore the influence of sentiment words that have no sentimental color on micro-blog sentiment, and lack of analysis for complex sentence. To solve the problem, this paper proposed a method of cross-grained sentiment analysis based on conditional random filed and complex sentence, which fused complex sentence and semantic dependency features on the basis of CRF. It can identify sentiment elements by analyzing micro-blog sentiment in fine-grained. The method of coarse-grained sentiment analysis based on complex sentence was used to analyze sentimental tendency of student micro-blog. Finally, the experimental results show that the accuracy on sentiment elements can reach 88%, furthermore, the accuracy of micro-blog sentimental tendency can reach 87%. Compare to traditional method, the method we proposed has higher accuracy and better performance.