【作者】 张云中；

【导师】 徐宝祥；

【作者基本信息】 吉林大学 ， 情报学， 2012， 博士

【摘要】 Web向社会化与语义化的不断演进和信息资源组织理论的不断革新共同促生了folksonomy并推动其不断发展，而folksonomy的不断优化又离不开folksonomy知识发现理论的支撑，同时形式概念分析、本体等理论的发展又为folksonomy知识发现注入了新的活力，一种基于形式概念分析的folksonomy知识发现理论呼之欲出！回顾当前folksonomy知识发现理论的优势劣态，虽在folksonomy知识发现的各个主要方向取得了一些散落的成果，体现于folksonomy用户行为、folksonomy用户偏好和folksonomy语义关系等方面，但仍未建立完善的将三者统一到一套完整框架下的Folksonomy知识发现理论体系，更缺少对对Folksonomy知识发现的基本原理、基本方向、目标产物、技术工具和详细流程等全面介绍与阐述。基于形式概念分析的folksonomy知识发现理论为弥补上述缺陷提供了可能。在确立folksonomy是数据，形式概念分析是工具，知识发现是目标的角色定位后，基于形式概念分析的folksonomy知识发现螺旋演进模型应运而生，其高度概括了基于形式概念分析的folksonomy知识发现的组成要素、角色要素、功能要素及各个要素之间的紧密关系，并将基于形式概念分析的folksonomy知识发现过程归纳为问题定义、数据获取、数据准备、数据组织、数据挖掘、知识生成和评估反馈七个阶段。另外，在用户需求和数据组织的一唱一和下，基于形式概念分析的folksonomy知识发现核心方向的歧化也通过folksonomy多值形式背景的选择得以实现，folksonomy的用户行为背景、用户偏好背景和语义关系背景分别决定了基于形式概念分析的folksonomy知识发现的三大方向。基于形式概念分析的folksonomy用户行为分析以单用户的用户标记行为、用户群的聚集与形成、用户群标记行为和用户群的遴选为知识发现目标，通过基于形式概念分析的folksonomy用户行为分析模型，先利用folksonomy数据集构建相应的folksonomy用户行为形式背景——“用户-标签”形式背景FU:= (U,T×R,YU)并将其转换为folksonomy用户行为概念格，并在folksonomy用户行为概念格分析的基础上使用回溯法获取folksonomy单用户用户标记行为链，使用用户群层级结构映射规则获取folksonomy用户群层次树，使用FREtirj计算公式获取folksonomy用户群标记行为频率，使用folksonomy用户群遴选的规则遴选出folksonomy典型用户群。folksonomy用户行为分析为folksonomy用户偏好挖掘提供活跃用户、典型用户群及folksonomy用户标记行为频率等参数，为folksonomy语义关系发现提供低频标签过滤、稳态folksonomy系统判断及语义浮出判断的依据。基于形式概念分析的folksonomy用户偏好挖掘以folksonomy用户偏好树的构建为目标，通过基于形式概念分析的folksonomy用户偏好挖掘模型，分别从活跃单用户和典型用户群的用户偏好各自的数据集出发构建相应的folksonomy用户偏好形式背景——“资源-用户”形式背景FR:= (R,U×T,Y~R)，并分别将单用户folksonomy用户偏好形式背景和用户群folksonomy用户偏好形式背景转化为各自的用户偏好概念格。通过在概念格基础上识别用户偏好，并借鉴TF/IDF原理充分考虑了用户偏好的“频率”和“普遍重要性”两项因素，分别提出了folksonomy单用户偏好权重计算公式和folksonomy用户群偏好权重计算公式，进而分别利用相应的用户偏好权重计算公式（PW公式）和用户偏好相似度计算公式（PS公式）遴选单用户的用户偏好和用户群的用户偏好，最终构建folksonomy用户偏好树来表示folksonomy用户偏好知识。基于形式概念分析的folksonomy语义关系发现以发现folksonomy中的隐含语义为目标，确立了本体在folksonomy语义关系表示中的重要作用，通过基于形式概念分析的folksonomy语义关系发现模型，从“语义浮出”的稳态folksonomy的数据集出发构建相应的folksonomy语义关系形式背景——“资源-标签”形式背景FR:= (R,T×U,Y~R)，并将之转化为相应的概念格。在folksonomy语义关系概念格基础上，利用folksonomy语义关系概念格向folksonomy局部本体的映射规则，得出相应的folksonomy局部本体模型，之后选用合适的本体编辑工具（如protégé等）和本体描述语言（如owl语言）对folksonomy局部本体模型进行形式化，最终得到一个形式化的揭示folksonomy各种隐含语义关系的folksonomy局部本体。基于形式概念分析的folksonomy知识发现的理论体系充分利用了folksonomy多值形式背景的不同形式适宜地表示了folksonomy知识发现的三大核心方向，并针对不同类型的概念格分别提出了相应的模式识别和知识解释方式，从而依托概念格实现了folksonomy知识发现。另外，基于形式概念分析的folksonomy知识发现的螺旋演进思想也为获取用户满意的folksonomy知识提供了保障。经过在Delicious系统数据的测试，基于形式概念分析的folksonomy知识发现理论具有创新性、科学性、合理性和操作性。基于形式概念分析的folksonomy知识发现理论不仅深化了folksonomy理论研究，也拓展了知识发现理论的内涵和外延，最重要的是揭示了基于形式概念分析的folksonomy知识发现的客观规律。该理论必将提高folksonomy系统中知识发现的效率和能力，进而带来folksonomy自身的不断优化，最终促进web2.0下folksonomy的不断发展和广泛应用。因此，无论是理论上还是实践上，基于形式概念分析的folksonomy知识发现理论都具深远意义！更多还原

【Abstract】 Both the constantly evolving of web to socialization and semantization andthe constant innovation of information resources organization theory not onlyfacilitate the emergence of folksonomy but also promote its continuousdevelopment. The optimization of folksonomy can not be separated from thesupport of folksonomy knowledge discovery, while formal concept analysis,ontology and other related theory has injected new vitality for folksonomyknowledge discovery, A FCA-based folksonomy knowledge discovery theoryget ready to come out!when reviewing the advantages and disadvantages of knowledgediscovery theory in folksonomy，it can be found that some scattered results inthis area has been achieved, including user behaviors, user preferences andsemantic relations in folksonomy, but a prefect knowledge discoverytheoretical system which integrate the three directions into a completeframework has not yet been established in folksonomy, along with the lack of acomprehensive introduction to the basic principles, basic direction, goals andproducts, technology and tools, detailed operational processes of knowledgediscovery in folksonomy.The FCA-based folksonomy knowledge discovery theory offers thepossibility to compensate for these shortcomings. After the respective role offolksonomy, FCA and knowledge discovery has been determined, AFCA-based folksonomy knowledge discovery spiral evolution model came intobeing. the model highly summarizes the constituent elements，role elements，functional elements and the tight relationship between these various elementsin FCA-based folksonomy knowledge discovery，and divides the wholeprocess into seven stages which are orderly composed of problem definition,data acquisition, data preparation, data organization, data mining, knowledge generation and evaluation&feedback. In addition, with the user needs and dataorganization echoing each other, the disproportionation of the core direction inFCA-based folksonomy knowledge discovery is implemented by choosingappropriate multi-value context of folksonomy. In deed, the user behaviorcontext, the user preferences context and the semantic relations contextrespectively decide three major directions of FCA-based folksonomyknowledge discovery.FCA-based folksonomy user behavior analysis takes tagging behavior ofsingle-user, gathering and formation of user group, tagging behavior of usergroup and typical user group selection as the targets. With the supporting ofFCA-based folksonomy user behavior analysis model, A folksonomy userbehavior formal context called“user-tag”context which denoted as FU:=(U,T×R, Y~U) is constructed from folsonomy data sets, and then the context isconverted to the folksonomy user behavior concept lattice. on the basis ofanalyzing the user behavior concept lattice, the single-user tagging behaviorchain can be obtained by retrospective method, the user groups hierarchy treecan be got by user group hierarchy mapping rules, the frequency of taggingbehavior of user groups can be calculated by the FREtirj formula, also thetypical user group can be obtained by selection rules. FCA-based folksonomyuser behavior analysis not only provides the parameters just as activesingle-user, typical user groups and the frequency of tagging behavior forfolksonomy user preferences, but also provides the basis for low-frequency tagfiltering, steady-state folksonomy system judgment and semantics emergingjudgment in folksonomy semantic relations discovery.FCA-based folksonomy user preferences mining take the construction ofuser preferences tree as the target. under the supporting of FCA-basedfolksonomy user preferences mining model,and beginning with respective datasets of active single-user and typical user group, folksonomy userpreferences formal context called“resource-user”context which denoted asFR:= (R,U×T, Y~R) is constructed, and then the context for single-user and user group is converted to the folksonomy for each other. Then the user preferenceweights formula for the two is proposed by identifying the user preferences onthe user preferences concept lattice and considering the factors of“frequence”and“universal importance”which learned from the TF/IDF principle. Finally,the folksonomy user preferences tree is build for user preferences expressionthrough sorting user preferences according to user preferences weight formulaand user preferences similarity formula.FCA-based folksonomy semantic relations discovery takes the impliedsemantic in folksonomy as the target and recognize the important role ofontology in the expression of folksonomy semantic relationship. Through thesupporting of FCA-based folksonomy semantic relations discovery model, Afolksonomy semantic relations formal context called“resource-tag”contextwhich denoted as FU:= (U,T×R, Y~U) is constructed from a“steady–state”and“semantics emerging”folsonomy data sets and then converted to thefolksonomy semantic relations concept lattice. Using mapping rules fromfolksonomy semantic concept lattice to the local ontology, the local ontologymodel is constructed on the basis of analyzing the folksonomy semanticconcept lattice. Ultimately, the formal local ontology for revealing variety ofimplicit semantic relations in folksonomy is achieved by choosing appropriateontology editing tools (such as the protégé) and ontology description language(such as the owl language) to formalize the local ontology model.The FCA-based folksonomy knowledge discovery theoretical system is anew and innovative theory that makes full use of the different forms of thefolksonomy multi-value contexts which befittingly show the three core directionof folksonomy knowledge discovery, and propose corresponding patternrecognition and knowledge explain methods for different types of conceptlattice, thus realize knowledge discovery relyingon the concept lattice. Inaddition, the idea of spiral evolution in FCA-based folksonomy knowledgediscovery process also provided a guarantee to obtain user-satisfiedfolksonomy knowledge. Finally, we can get the conclusion that the new FCA-based folksonomy knowledge discovery theory is scientific, reasonableand operational after the data test using dataes from Delicious web site.The FCA-based folksonomy knowledge discovery theory not onlydeepens the folksonomy theoretical study, but also extends the intent andextent of the knowledge discovery theory, most importantly, it reveal theobjective laws of FCA-based folksonomy knowledge discovery.the theory willimprove the efficiency and capacity of knowledge discovery in folksonomysystems, promote folksonomy constantly self-optimization, and ultimatelyboost the continuous development and wide application of folksonomy inweb2.0 environment. Therefore, both in theory and in practice, FCA-basedfolksonomy knowledge discovery theory has far-reaching significance!更多还原