中文摘要：

为解决日益严重的垃圾博客问题，对产生垃圾博客的作弊技术和相应的识别技术进行了研究。通过对大量中文垃圾博客的分析，结合对作弊者目的的研究，提出了从用户名、发帖时间间隔、博文内容、锚文本和链接地址、分类标签等博客的结构特征出发的特征提取方法。在特征提取的基础上，提出了基于多结构特征的识别方法，并建立了相应的系统模型。使用支持向量机和朴素贝叶斯模型作为分类器进行了实验，并与经典的基于内容的方法进行了对比。实验结果表明，在小的训练集上，基于多结构特征的方法正确率达到90％以上，比基于内容的方法提高了6个百分点，该方法可有效区分垃圾博客和正常博客。

英文摘要：

To address the growing problem of Splog, the generating Splog technology and the corresponding recognition technology are studied. By analyzing a large number of Chinese Splog and the purposes of Splog maker, a method of extracting feature from blog structure features is proposed such as the user＇ s name, post time interval, post content, anchor text and link address, classification labels. Based on the feature extraction, a method based on the multiple structure features is proposed. The naive Bayesian model and support vector machines are used as the classifier in our model. Experiments on a small train dataset show that the method based on multiple structure features reaches an accuracy of 90%. Compared with the contend based method, proposed method increases the accuracy by 6%, indicating that the method can identify Splogs effectively.