中文摘要：

指出微博在传播信息的同时，也夹杂着谣言等虚假消息、不实言论。针对微博谣言传播速度快、影响范围广等特点，深层挖掘微博中的隐含信息，提出符号特征、链接特征、关键词分布特征和时间差等新特征，将微博谣言识别形式化为分类问题，综合新提取的特征与微博文本特征、用户特征和传播特征构建多个特征模板，利用SVM分类学习方法对微博进行分类，识别结果可有效辅助人们更好、更快地识别谣言。实验结果表明，在基本特征的基础之上，新提出的特征能有效提高微博谣言识别的正确率。

英文摘要：

Microblog not only disseminates information, but also is mingled with rumors and false news. In view of microblog rumors rapidly spreading with wide scope of influence, new features such as symbol, links, keywords distribution and delta - T are proposed by deeply mining the feature information implied in microblog. Rumor identification is formulated as classification problem. Different feature templates are built with new proposed features and classic features like text features, user features and propagation features of microblog. Then SVM is used to classify microblog to help effectively identify rumors. The experimental results suggest that the new features proposed based on the basic ones significantly promotes the overall accuracy of rumor identification.