中文微博的热点话题发现

发布时间：2018-03-01 04:31

本文关键词： 微博热点话题文本聚类关键词　出处：《安徽大学》2014年硕士论文　论文类型：学位论文

【摘要】：近些年,微博已然成为社会网络一个很重要的的应用,人们通过微博可以在任何地点,任何时间发布自己的所见,所闻,所想。这种较为简易的消息发布方式很大程度上降低了信息发布的壁垒,这也是微博得到广泛应用的重要因素之一。在这些用户发布的微博数据中,多数的微博信息都和现实生活中的事件有所关联。有些微博网站也会在网站主页中列出最近一段时间的热门关键词,但是这些微博主题词通常都很简短、分散,而且与某个话题有关的微博信息是由不同微博发布者在不同时间内发布的。用户不能通过这些孤立信息对事件有一个全面的了解,相关部门也不能完整掌握事态的发展,对微博上的舆论不能进行正确的引导。正是在这种情形下,针对微博数据的热点话题发现逐渐成为一个研究热点。微博热点话题发现就是通过相关的话题检测技术把分散的微博信息组织起来有条理地呈现给人们。本文提出了一个完整的微博热点话题发现方法。该方法主要包括了微博数据的预处理、微博数据的文本表示、特征词提取、文本聚类和事件关键词提取等步骤。文中仔细分析了微博数据自身的特点,结合这些微博数据特征对TF-IDF算法进行改进,提高了微博热点话题发现的效果。本文在微博文本聚类时采用了多属性无向加权图聚类算法,不仅考虑了图的拓扑结构还引入了节点的属性,与目前常用的一些聚类算法相比较提高了聚类的效果。文中给出的一些实验结果表明本文提出的微博热点事件检测算法是有效的。
[Abstract]:In recent years, Weibo has become a very important application of the social network. People can publish what they see and hear from any place, any time, through Weibo. Think about it. This relatively simple way of publishing information has greatly reduced the barrier to information release, which is also one of the important factors that Weibo has been widely used. In the Weibo data published by these users, Most Weibo information has something to do with real-life events. Some Weibo websites also list popular keywords from recent periods of time on their home pages, but they are usually short and scattered. Moreover, Weibo information related to a certain topic was published by different Weibo publishers at different times. Users cannot have a comprehensive understanding of the incident through these isolated information, nor can the relevant departments fully grasp the development of the situation. It is under such circumstances that the public opinion on Weibo cannot be properly guided. The hot topic discovery of Weibo data has gradually become a research hotspot. Weibo hot topic discovery is to organize and present the scattered Weibo information to people in a coherent way through the relevant topic detection technology. In this paper, a complete method of hot topic discovery by Weibo is proposed. This method mainly includes the pretreatment of Weibo data, the text representation of Weibo data, the extraction of feature words. Text clustering and event keyword extraction. This paper carefully analyzes the characteristics of Weibo data itself, and improves the TF-IDF algorithm by combining the characteristics of the Weibo data. In this paper, we adopt multi-attribute undirected weighted graph clustering algorithm, which not only considers the topological structure of graph, but also introduces the attributes of nodes. Compared with some commonly used clustering algorithms, the clustering effect is improved. Some experimental results show that the Weibo hot spot event detection algorithm proposed in this paper is effective.
【学位授予单位】：安徽大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.1;TP393.092

【相似文献】