基于微博内容的博主标签生成关键技术研究

发布时间：2018-04-10 16:11

本文选题：微博 + 用户标签　；参考：《国防科学技术大学》2014年硕士论文

【摘要】：近几年,以Twitter、新浪微博为代表的微博客(Micro Blog)应用得到了快速的发展。吸引了大量的用户的同时,积累了海量的用户数据。已渐渐的发展成了一种被大众所喜爱的新型媒体。从用户数据中挖掘出用户特征,为微博用户提供个性化的服务,为企业和政府部门提供相关建议,这使得该项研究有着巨大的商业价值和社会价值。本文以微博用户发表的微博内容作为输入数据,运用微博关键词提取、微博用户建模和分类技术,发现微博用户的特征,为微博用户添加一个描述用户特征的标签。本文对关键词提取、用户建模和分类技术进行了深入的研究,主要的工作如下:首先,在微博关键词的提取方面,根据微博短文本的特性,运用TF-IDF算法和Text Rank算法计算词语在微博中的权重;然后将用户的微博转化成空间向量模型,使用聚类算法提取候选关键词;得到候选关键词后,运用n元语法模型对候选关键词扩展;最后根据邻接变化数和语义单元数对扩展后的关键词进行筛选,得到最终的关键词列表。有效的提取出了用户微博中的关键词。然后,在微博用户建模方面,将微博分成原创微博、转发微博、话题微博、主题微博。以用户添加过标签的微博为参照,研究不同微博在表现用户特征方面的不同。结果表明话题微博和主题微博更能够表现出微博用户的特征。基于此提出了一种基于微博类型的用户建模方案,加强了话题微博和主题微博在模型中的权重。有效的提升了建模效果。最后,在用户标签生成方面,以新浪微博平台的官方认证账号发表的微博为训练数据,利用本文的关键词提取方法和用户建模方案,为官方认证账号建立标签模型,为待添加标签用户建立用户模型。运用支持向量机(SVM)的分类功能,为用户生成标签。
[Abstract]:In recent years, Weibo Micro blog, represented by Twitterand Sina Weibo, has developed rapidly.Attract a large number of users at the same time, accumulated a huge amount of user data.Has gradually developed into a popular favorite of the new media.Mining user characteristics from user data, providing personalized services for Weibo users, and providing relevant suggestions for enterprises and government departments make the research have great commercial and social value.In this paper, we take the Weibo content published by Weibo user as the input data, extract the keywords of Weibo, and discover the user's characteristics by the techniques of user modeling and classification. Then we add a label describing the user's characteristics for the user.In this paper, keyword extraction, user modeling and classification techniques are deeply studied. The main work is as follows: firstly, in the aspect of keyword extraction of Weibo, according to the characteristics of the short text of Weibo,TF-IDF algorithm and Text Rank algorithm are used to calculate the weight of the words in Weibo. Then the user's Weibo is transformed into a spatial vector model, and the candidate keywords are extracted by clustering algorithm.Finally, the extended keywords are filtered according to the number of adjacent changes and the number of semantic units, and the final keyword list is obtained.Effective extraction of user Weibo in the key words.Then, in the Weibo user modeling aspect, the Weibo is divided into the original Weibo, forwards Weibo, the topic #china_person-, the theme Weibo.Taking Weibo, who has been tagged, as a reference, this paper studies the differences of user characteristics between the two groups.The results show that the topic Weibo and the theme Weibo are more able to show the characteristics of the user.Based on this, a user modeling scheme based on Weibo type is proposed, which strengthens the weight of the topic Weibo and the theme Weibo in the model.The modeling effect is improved effectively.Finally, in the aspect of user label generation, the official authentication account of Sina Weibo platform is used as training data, and the keyword extraction method and user modeling scheme are used to establish tag model for official authentication account.Create a user model for the user to be tagged.Support Vector Machine (SVM) is used to generate tags for users.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.092;TP391.1

【相似文献】