如何从跨网页多源异构数据中挖掘个体之间的非隐私真实社会关系，以构建真实的社会关系网络，是社会计算领域一个新的研究课题。区别于现有的实体关系抽取任务，跨网页多源异构数据社会关系挖掘将原有文本内部局部范围（句内或段内）的实体关系抽取扩展为跨文本跨网页中的实体抽取、同名消歧、关系识别和数据融合等任务，具有一定的挑战性。课题融逆向最大匹配、HMM模型和人名可信度模型对文本分词、词性标注和人名识别等任务进行联合协同分析处理，用个人关键特征信息相似度计算方法实现同名消歧；在人名识别和同名消歧基础上，用语义规则、关系模式匹配、关系自动分类相结合的方法，以及用搜索引擎、关系自动分类、融合选举投票思想等两种方案来实现跨网页社会关系挖掘。课题研究为构建真实社会关系网络提供了途径，对社会关系和社会网络的研究具有重要意义，构建的真实社会关系网络对于评审回避和面试回避等相关工作具有重要的实用价值。

It is a new research in the field of social computing to establish a real social relations network on how to mine the non-privacy real social relations between individuals from the cross-Webpage Multi-source heterogeneous data. Different from the existing entity relation extraction task, the social relations mining of the cross-Webpage Multi-source heterogeneous data extend the existing entity relation mining of internal partial range of the text (sentence or paragraph) to the task of entity extraction, disambiguation of the same name, relationship recognition, data fusion and so on in the cross-text or cross-Webpage data, which is a great challenge to the research. The project combines the Reverse Maximum Matching, the Hidden Markov Model and the Names Credibility Model to do with the tasks of word segmentation, POS tagging, name recognition and so on in a joint collaborative analysis way. And it uses a similarity calculation method of the personal key feature to achieve disambiguation of the same name. Based on the above name recognition and the disambiguation of the same name, those can help achieve cross-Webpage social relations mining by using a combination method of semantic rules, relations pattern matching and automatic relations classification, and another method of search engine-based, automatic relations classification, fusion elections thinking. The research provides a way to establish a real social network, which has great significance for the study of the social relations and social networks. Besides, the established real social network has much practical value for assessment evasive, interview evasive and the related work.