基于语义空间的藏文微博情感分析方法

藏语微博是目前流行的藏文网络媒体形式。对藏文微博文本进行情感挖掘，能够有效提高政府对藏语言的监测能力。传统的文本分类方法对中文微博能够达到不错的效果，但由于藏文具有自身的语言特点，传统方法对藏语的分类效率并不高。本文提出了一种基于语义空间的藏文微博情感分析方法。该方法首先使用句法树生成句法结构；然后结合句法结构和语义特征向量构建语义特征空间，在特征空间中通过K-means方法聚类形成语义簇质心；最后计算基于簇的TF-IDF值作为最终的微博情感特征值。实验结果表明，与目前常用的SVM TF-IDF和Naive Bayes 最大熵方法相比，该方法能更准确地对藏文微博进行情感分类。

Emotional Classification Method of Tibetan Micro-blog Based on the Semantic Space

Tibetan Micro-blog is a popular form of Tibetan online media. Mining Tibetan Emotional can effectively improve the ability of the government to monitor the Tibetan language.To mine Chinese Weibo text, traditional classification method can achieve good results witch for Tibetan classification efficiency is not better. This paper presents an emotional classification method of Tibetan Micro-blog that based on the semantic space. Firstly, the syntactic structure is generated using the syntax tree; Then combined syntactic structure and semantic feature vector to construct the semantic feature space. In the feature space,it will form semantic cluster centroid method by K-means clustering; Finally,the emotional values of Micro-blog are calculated by TF-IDF witch base on the clusters. Experimental results show that this method can be more accurately classified on Tibetan Micro-Blog text emotion, compare with SVM TFI-DF and Naive Bayes maximum entropy witch are the most commonly used method.