摘要
针对高维、稀疏的中文微博数据,提出一种多步骤的新闻话题发现方法。首先结合微博的传播特点,选取出不同时间窗口中具有较高新闻价值的微博文本;再利用隐主题模型挖掘微博内容中隐含的主题信息,并在此基础上进行文本聚类;最后使用频繁项集挖掘技术获取话题关键词集合。该算法能够较好地实现对中文微博数据的降维与话题发现。真实的微博数据集实验结果验证了该方法的有效性。
For high dimensional and sparse Chinese microblog data, this paper proposed a multi-step method for discovering topic. Firstly, it combined with the spread characteristics of microblog, it got the microblog content which had a high news va- lue. Then, it used the hidden topics analysis technique to model the text data and got the result of the text clustering by using the hidden topic information. Finally, the keywords which were best represented the topic content would be obtained from the clustered results through frequent itemsets mining. The experimental results verify the validity of the method on Chinese mi- croblog dataset＇ s dimensionality reduction and topic identification.
出处
《计算机应用研究》
CSCD
北大核心
2014年第3期700-704,共5页
Application Research of Computers
基金
国家科技支撑计划课题资助项目(2012BAH18B05)
四川大学青年教师科研启动基金资助项目(2013SCU11017)
关键词
中文微博
话题发现
隐主题模型
文本聚类
频繁项集挖掘
Chinese microblog
topic discovering
hidden topic analysis
text clustering
frequent itemsets mining
