导　　师： 杨建梅;徐学军

学科专业： 1201

授予学位： 博士

作　　者： ;

机构地区： 华南理工大学

摘　　要： 在线百科是典型的大众生产社区。本文基于网络谱方法探讨百度百科社区上大众生产者（贡献者）的异质性协作关系特征。具体工作体现在如下几个方面： 根据“临界质量”理论的基本观点分析了社区贡献者的异质性以及参与特征。研究表明，社区内的贡献者在资源和“利益”方面都呈现出异质性。社区内容生产函数更多地属于加速生成函数。因此，百度百科社区满足内容产生并可持续发展的关键条件。此外，社区贡献者不论是“局部”贡献值，还是“整体”贡献值的分布都明显服从“少数”定律。不仅如此，由这些贡献者参与行为导致的社区内容的“严谨性”和“多样性”方面同样服从这种特征。 提出采用反映贡献者异质性的网络分析方法研究贡献者的协作关系。研究表明，贡献者协作网络属于无标度异配型高聚集性网络。在知识专业性比较强的社区内，贡献者的协作网络特征与其在社区内的贡献值具有显著的相关性。反之，在知识专业性比较宽泛的社区内却不存在这种相关性。在采用动态网络分析相关性时发现，贡献者在网络中的连接强度、介数都始终与其编辑词条数存在较为显著的正相关性；而聚类系数则与之呈现出显著的负相关性。特别地，贡献者的接近中心性与其贡献值的相关性随着时间的演化出现明显的逆转现象。 采用网络谱的分析方法研究了由网络结构导致的集体状态以及网络上的局域性特征。基于贡献者异质性，通过引入二次量子化算子表述下的分子系统Hückel模型，建立了一般网络谱和对应量子系统能量谱之间的关联性，拓展了网络谱的研究方法。研究结果表明：百度百科社区贡献者动态协作各截面网络导致的集体状态中，部分属于集体有序态，部分属于集体软混沌态。在紧束缚近似下，随着时间演化各截面网络上对应的电子运动状态愈发趋向于完全局域态，且在演化中电子集体运动状态与其阶梯函数状态的偏差呈现出近似周期性波动。采用谱间距序列分析同样显示，部分截面网络对应的分子系统中电子集体运动的状态显著地处于局域态，而部分还处于中间态。 采用时间序列分析方法分别研究网络谱的波动序列和趋势序列。研究发现，贡献者动态协作关系在不同时间截面上网络谱的波动序列特征满足幂律关系，且这种幂律关系服从量子系统中能量谱波动序列特征的普适规律。同时，贡献者动态协作关系网络各截面网络谱序列都具有多重分形结构特征。 总之，本文研究了百度百科社区贡献者的协作关系的结构特征，提出反映个体异质性的一般网络的量子映射模型，建立了网络谱和量子系统能量谱的关联性。因此，既有实证研究的价值，同时还具有方法论的价值。 Online Encyclopedia is a typical community of the Peer Production. The characteristics ofcollaborative relations among heterogeneous producters are studied by using the networkspectrum methods in Baidu Encyclopedia. The specific contents as following: The contributors’ heterogeneity and behavioral characteristics are studied by introducingthe basic concepts of Critical Mass Theory. The results show contributors heterogeneity notonly presents in resource, but also in interest. Content production function is more likeaccelerating function. Therefore, Baidu Encyclopedia has the critical conditions that thecollective content are started and become self-sustaining. Besides, both of the contributor’slocal and whole contributions satisfy the collective phenomenon in Critical Mass Theory, i.e.,a small segment of the population that chooses to make big contributions to the articles whilethe majorities do little or nothing. The same phenomenon also exists on the “Rigor” and“Diversity” of these articles. The network analysis method which reflects the diversity of contributors is broughtforward to study the characteristics of collaborative relations among the contributors. Theresults show that collaborative networks are scale-free and disassortative network with highaggregation. There exists a significant correlation between the node characteristics andcontribution in community with relatively strong professional knowledge, but not incommunity with relatively broad professional knowledge. Furthermore, the results oncorrelations in dynamic network analysis show that Connectivity and Betweenness alwayshave s significant positive correlation with user’s contributions while Clustering always havea significant negative correlation with contributions. Particularly, there exists a reversal overtime in correlation between Closeness and contribution. The collective state induced by structure of dynamic collaborative network and thelocalization on dynamic collaborative network are studied by employing the networkspectrum methods. Due to the conctributor heterogeneity, we create the relation betweennetwork spectrum and energy spectrum of corresponding system by introducing Hückelmodel of Molecular Systems with the second quantized approach. This mapping model extends the research method of network spectrum. The research results show that part of thecollective states induced by structure of dynamic collaborative network are collective order,and parts are collective soft chaos. In the tight-binding approximation system, the electronicmotions on sectional network with over time increasingly tend to strongly localization. Also,the collective motion state follows an an approximate cyclical fluctuations to deviate thestep-like function state. At the same time, the results through level spacing sequence ofspectrum show that part of the collective motions of ecectrons on sectional collaborativenetworks are in localized state while some other may in intermediate state. The characteristics of fluctuation sequence and trend sequence are studied by using thetime series analysis method. The results show that the Power Spectrum of fluctuationsequence of network spectrum in every sectional network of dynamic collaborative followsthe Power-law relation, which is in agreement with the universality in quantum energyspectrum fluctuation. Besides, all of the spectrum sequences of dynamic collaborativenetwork section have the multi-fractal structure. To sum up, this thesis studies the characteristics of collaborative relation, and bringsforward the quantum mapping model of weighted network reflecting the diversity ofcontributors to create the relation between the network spectrum and network spectrum.Therefore, this research not only has the value of empirical research, but also the value ofmethodology.

关 键 词： 大众生产 百度百科 贡献者异质性 网络谱 量子映射

分 类 号： [F49 F224]

领　　域： [经济管理] [经济管理]