基于多社交媒体的个体身份关键技术研究

Key Technology Research on User Identity Resolution Across Multi-social Media

随着互联网的发展以及社交网络服务的普及,人们使用互联网传播信息的门槛逐渐降低,越来越多的人们通过多种多样的社交媒体来分享自己的个人信息.如今的互联网,充斥着巨量的个人信息,包括用户关注的事物、生活中的新闻、用户对社会事件发表的评论等.且可以发现一个现象,这些个人信息并不是随意的分布在互联网中的,而是有规律的按照社交媒体的特质分门别类的分布在各个社交媒体之中.所以通过任何一个单一的社交媒体了解用户都只能获取该用户一部分的信息,只有通过多个社交媒体共同的去了解用户,才能得到较为完整的信息.在不同的社交媒体之中发现一个人的多个社交账号(亦即个体身份识别)是社交数据融合的前提.个体身份识别是一个从不同的数据来源识别单一个体的过程.提出一种方法使用用户社会网络关系以及用户行为模式进行个体身份匹配.在收集了可能获得的用户社交网络以及用户信息之后,这个方法对这些信息进行分析和比较,包括进行网络之中结点的相似性计算以及字符串相似性计算,最后根据计算结果判定是否匹配.该方法划分为两个模块,分别为相似度计算模块与计算结果优化模块.

With the development of the Internet and the prevalence of SNS （ social networking service ）, the threshold of people using Internet to spread information becomes lower,and people increasingly share their personal information through kinds of social media. Nowadays, Internet is fulfilled by huge-scale personal information, which includes the interesting things people followed, daily news, people＇s comments towards social events,etc. There exists a phenomenon that these kinds of personal information are not distributed in the Internet randomly,but archived in several social media under a certain rule,which is said the different characters of social media. It is rational to infer to the conclusion that we could not learn the every aspect of an individual through only one social medium he/she using, and only if through multi-social media could we learn the more integrated information of this individual. Discovering multiple accounts of a single person across different social media is the precondition of social data fusion. User identity resolution is the process of identifying single user from disparate data sources. This thesis proposes a approach for user identity resolution based on user＇s social network and user＇s behavior patterns. After collecting available user＇s data, the approach accomplishes this task by similarity measure between vertices of networks and usernames, and results evaluation and analysis. This system consists of two modules, namely similarity calculation module and calculation result optimization module.