【作者】 赵鹏；

【导师】 蔡庆生；

【作者基本信息】 中国科学技术大学 ， 计算机应用技术， 2006， 博士

【摘要】 继二十世纪末复杂网络的小世界效应及无标度性的发现之后，复杂网络的研究得到了越来越多的关注，来自各个学科的研究者们从各个层面对复杂网络展开了深入的研究，复杂网络已经成为一个充满生命力的交叉研究领域。目前，复杂网络的研究主要集中在两个方面：一方面是复杂网络理论性的分析与仿真，新的理论模型和新的分析方法不断涌现；另一方面是从现实网络中不断发现新结构与新现象，运用复杂网络理论来观察、理解和解决具体应用问题。 随着信息技术的发展和互联网的普及，Web2.0已经成为新一代互联网应用的发展趋势。Web2.0系统中存在着大量的非线性、自组织和涌现现象。将复杂网络研究与Web2.0相结合，不仅有助于正确认识和理解Web2.0，以及对Web2.0的下一步发展有指导性意义，同时也将启发、推动复杂网络的理论研究工作。 个性化是Web2.0的主要特点之一。个性化信息服务已经成为互联网应用的一个重要的研究热点，得到了越来越多研究者的关注。其中，用户建模，聚类、分类以及自动推荐技术又是个性化信息服务中的关键技术，这些技术的研究必将有力地推动互联网大规模的个性化信息服务。 本论文围绕以上几个方面，将复杂网络的理论方法与互联网个性化信息服务相结合，进行了深入的研究和实践。论文的主要内容为： 首先，将复杂网络的研究与Web2.0相结合。具体包括：一、用复杂网络的理论方法，研究Web2.0系统中存在的非线性机制，自组织和涌现现象。二、研究复杂网络中的社团(community)发现理论，提出具有交联结构的复杂网络中的可重叠社团发现算法。三、对Web2.0中一个具有交联结构的复杂网络中的社团结构进行统计分析。 其次，将复杂网络特征应用于关键词抽取和聚类分析中，具体包括：一、研究了汉语语言所组成的自然语言网络中的“小世界”特性，提出基于复杂网络特征的关键词抽取算法。该算法综合考察单词在语言网络中的连接度和聚集性质，抽取复杂网络综合特征值高的节点作为关键词，旨在找到那些可能相对低频，但对文章主题表达起重要作用的单词。二、在对复杂网络重要特征深入研究的基础更多还原

【Abstract】 After the discovery of small-world phenomena and scale-free characteristics of complex networks in the late 20^th century, the research of complex networks has been gotten more and more attention. Researchers from different fields studied complex networks from every level. Complex networks has become a vital topic crossed with many other research fields. Nowadays the research of complex networks has been mainly focused on the two aspects: One aspect is theoretical analysis and simulation, in which new theoretical models and methods were proposed continually; The other aspect is applied research, in which new structures and phenomena of real-world network were discovered continually.With the development of information technology and the popularity of Internet, Web2.0 has become an important trend in the application of Internet. There are many non-linear, self-organize, and emergence phenomena in Web2.0 systems. So it is important that the theoretical methods of complex networks are applied to Web2.0, which can not only benefit understanding Web2.0 and guide the further development of Web2.0, but also can accelerate the theoretical research of complex networks.Personalization is a major feature of Web2.0. Personalized information service has been one of the hottest research points in the applied research of Internet. User profiling, clustering, classification and automatic recommendation are the crucial techniques in personalized information service. So the research of these crucial techniques will promote the large-scale personalized information service on Internet efficiently.This dissertation focuses on the above aspects and combines the theoretical methods of complex networks with the research of personalized information service. The research work of this dissertation can be summarized as follows:Firstly, this dissertation combines the research of complex networks with Web2.0. Main works are summarized as follows: （1） The non-linear mechanism, self-organize and emergence phenomena in Web2.0 systems are studied using the theoretical methods of complex networks. （2） A novel algorithm for finding the overlapping community in the complex networks with intersection structure is proposed. （3） The overlapping community structure in the complex networks with intersection structure is analyzed statistically.Secondly, the features of complex networks are applied to the automatic keywords extraction and clustering analysis. Detailed works are summarized as follows:（l） The small-world structure in human language network of Chinese is studied. A novel automatic keywords extracting algorithm based on the features of complex networks is proposed. This algorithm extracts those words with higher degree and clustering coefficient in the language network as keywords. Its goal is to extract keywords which may be relatively low frequency, but do great contribute to the subject of the text. （2） The definitions of the weighted complex networks features are given after the deeply studying on the features of complex networks. A novel更多还原