基于异质网络时态语义路径相似度的人物校正方法

发布时间：2018-06-24 12:03

本文选题：异质关系网络 + 时态语义路径　；参考：《中南大学》2014年硕士论文

【摘要】：社会关系网络分析已经成为当今最热门的研究领域之一。在社会关系网络建立的过程中,由于数据来源的多样化、数据规模的日益扩大以及事务信息的不完备、人物的个人基本信息缺失、信息模式与结构的差异以及曾命名现象等因素,导致关系网络中人物关系出现混乱,这将使得在社会关系网络中人物的唯一性难以识别。基于普通关系网络结构参数计算的传统社会关系网络分析方法难以解决上述问题。本文在异质关系网络以及语义网络基础之上,提出了一种基于异质关系网络的时态语义路径相似度计算的人物唯一性度量方法以及基于结构误差计算的相同人物合并校验策略。本文调研了传统社会关系网络分析中常用的普通关系网络模型以及异质关系网络模型,分析了普通关系网络在大规模社会关系网络分析中的局限性,提出了基于异质关系网络的时态语义路径相似度的人物唯一性度量方法。该方法结合语义网络的特征提出了时态语义网络的概念,利用人物节点对之间的时态语义路径相似度计算以及基于相似度阈值的唯一性人物过滤策略来实现人物唯一性识别,通过包含了多样性的关系语义信息的相似度计算实现了人物节点相似性及唯一性的度量,实验结果表明该方法能够在大规模社会关系网络中准确度量人物节点的唯一性,验证了方法的有效性。针对关系网络中具有高相似度的人物节点,本文提出了关系网络结构误差计算方法,该方法根据合并前后人物节点的度的变化计算节点对的结构误差,根据结构误差的取值判断人物节点对在关系结构上是否完全相同。通过该方法将具有完全相同的关系结构的人物节点筛选出来。然后,用相同人物节点合并策略对其进行合并,从而实现了关系网络中人物校正的目的。本文通过对一个含有多种学术活动信息的学术关系网络数据进行实验分析,验证了该方法在关系网络中相同人物校正中的有效性。
[Abstract]:Social network analysis has become one of the most popular research fields. In the process of establishing social relations network, due to the diversification of data sources, the increasing expansion of data scale and incomplete transaction information, the basic personal information of people is missing. The difference of information pattern and structure, as well as the phenomenon of once naming, lead to the confusion of the relationship between people in the relationship network, which will make it difficult to identify the uniqueness of the person in the network of social relations. It is difficult to solve the above problems by traditional social network analysis method based on the calculation of common relational network structure parameters. On the basis of heterogeneous relation network and semantic network, this paper presents a new method to measure the similarity of temporal semantic path based on heterogeneous relation network and a new method of merging and checking the same person based on the calculation of structural error. This paper investigates the common relationship network model and heterogeneous relationship network model commonly used in the traditional social relationship network analysis, and analyzes the limitations of the common relationship network in the large-scale social network analysis. In this paper, we propose a method to measure the similarity of temporal semantic paths based on heterogeneous relational networks. Based on the features of semantic network, the concept of temporal semantic network is put forward in this method. The similarity calculation of temporal semantic path between human nodes and the unique character filtering strategy based on similarity threshold are used to realize the identity recognition. The similarity and uniqueness of human nodes are measured by the similarity calculation of relational semantic information including diversity. The experimental results show that the proposed method can measure the uniqueness of human nodes accurately in large-scale social networks. The validity of the method is verified. In this paper, a method of calculating the error of relational network structure is proposed for the person nodes with high similarity in the relational network. This method calculates the structural errors of the nodes according to the change of the degree of the nodes before and after the merging. According to the value of the structure error, it is determined whether the human nodal pair is exactly the same in the relation structure. By this method, the character nodes with exactly the same relationship structure are filtered out. Then, the same person node merging strategy is used to achieve the goal of character correction in the relational network. In this paper, the validity of this method in the correction of the same person in the relational network is verified by the experimental analysis of an academic relationship network data containing a variety of academic activity information.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09;TP391.1

【参考文献】