微博用户个性化标签提取技术研究

发布时间：2018-11-04 11:49

【摘要】：微博用户标签由于能反映用户特征,用户偏好等信息,并且用户标签对用户广告推荐,用户聚类,用户查找等都有一定潜在价值,因此研究微博用户标签提取有一定实际意义。本文研究的微博用户个性化标签中的个性化包含两层含义,一是标签能反应用户的个性化特征,二是标签自身包含相应的个性化特征。标签反应用户的个性化特征的程度本文通过与用户手动提取的标签对比进行判断,标签自身包含的个性化特征是指对标签进行的进一步分类处理,使用户的标签之间带有共同属性,以便于对用户进行查找,聚类等。本文通过研究发现用户自提标签中存在三种基础类型的标签,本文分别称为基础标签,分类标签,关注标签,之后分别研究了每种基础类型标签的特点,并根据每种基础类型标签的特点设计了相应的提取方法,之后又根据三种基础类型标签之间的关系,研究如何将三者综合到一起,得到更好的反应用户个性化特征的标签。因此本文对用户个性化标签的提取过程中共涉及到七种标签提取方法,其中三种方法提取的为基础类型的标签,另外四种方法则是这三种基础类型标签之间的混合标签,除基础类型标签中的基础标签的提取方法是使用的已有算法TextRank算法外,其余六种标签提取方法均是本文所提出。通过最后的验证实验发现,三种基础类型标签的混合标签提取效果最好,因此,本文所研究的用户标签提取方法使用户个性化标签提取的效果得到了改进。另外,本文对提取出来的用户个性化标签进行进一步分类处理之后,使微博用户的标签之间带有了更多共性的信息,对用户聚类,用户分类,用户查找等也带来了一定好处,使用户标签的应用范围变得更加广泛。
[Abstract]:Weibo user tags can reflect the characteristics of users, user preferences and other information, and user tags to the user advertising recommendation, user clustering, user search and so on have a certain potential value. In this paper, the personality of Weibo user personalized tag contains two meanings, one is that the tag can reflect the personalized characteristics of the user, the other is that the tag itself contains the corresponding personalized features. Tags reflect the degree of personalized features of the user this paper compares with the tags extracted by the user manually. The personalized features contained in the tags refer to the further classification of the tags. Make the user's tags with common attributes, to facilitate the user to find, clustering, and so on. In this paper, we find out that there are three basic types of tags in user self-removal tags, which are called basic label, classified label, concern label, and then the characteristics of each basic type tag are studied respectively. According to the characteristics of each basic type label, the corresponding extraction method is designed, and then according to the relationship between the three basic type tags, how to integrate the three labels together to get a better response to the personalized features of the user tags. Therefore, in this paper, there are seven kinds of tag extraction methods involved in the process of user personalized tag extraction, three of which are based on tags, the other four are mixed tags between these three basic types of tags. Except for the existing TextRank algorithm which is used to extract the basic label in the basic type tag, the other six label extraction methods are all proposed in this paper. Through the final verification experiment, it is found that the mixed tag extraction effect of the three basic types of tags is the best. Therefore, the user tag extraction method studied in this paper has improved the effect of user personalized tag extraction. In addition, after further classifying the extracted user personalized tags, this paper makes Weibo user tags with more common information, which also brings certain benefits to user clustering, user classification, user search, and so on. Make user label's application scope more extensive.
【学位授予单位】：哈尔滨工程大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP393.092;TP311.13

【相似文献】