基于向量空间模型的层次聚类算法在文本挖掘中的应用

在传统聚类模型的基础上,提出一种基于向量空间模型的层次聚类算法,用于文本数据的挖掘。实验结果表明,基于向量空间模型的层次聚类算法从挖掘的准确率上更具有性能优势。     

以研究方法创新促进考试研究的创新和发展

探讨研究方法的创新对考试研究的创新和发展的促进作用。对目前考试研究依据的理论和应用方法做出评述后,分别对多层线性分析、潜在类别分析和函数型数据分析这三种新型的数据分析方法的理论与应用作了阐述,同时对这些方法应用于考试研究的可能性及应用做了分析。并指出研究方法的进步是考试研究创新和发展的前提。     

基于高校扁平化管理的财务分级管理模式探讨

高校扩招后,高校管理工作面临着管理日益复杂、办学经费短缺、亟需提升内涵建设水平的新态势。因而高校管理应改变传统的“金字塔”式管理模式为“扁平化”的管理模式,高校财务管理模式也应由“集权式”转变为“分级式”的财务管理模式。     

网络资源层次组织规律定量研究

本文以互联网FTP目录树为对象,从大众构造的层次式结构中揭示管理网络资源时的组织规律。通过考察44,972个软件资源及其在目录树中的位置,得到如下认识:①对于用户查找、浏览资源来说,资源存放在深度为2或3层较适合。②当要管理的资源数量在200以内时,人们可容忍平均1个目录中包含10个左右的软件资源;当资源数量更大时,人们可容忍平均1个目录中包含15个左右的软件资源。③在使用层次式组织结构时,人们会对结构的深度有一个习惯上限,目录结构平均深度一般不多于6层。④人们考虑组织结构时,会在"分类体系复杂度"和"分类精细度"之间权衡,在资源量增多时,尽量保证扁形结构。本文揭示了采用层次方式组织资源时的一些共性,有助于设计更贴近大众习惯的网络资源组织与导航系统,并对标签聚类等从扁平型信息组织方式中提取类目层次的研究工作提供依据。     

Hierarchical Effects of School-, Classroom-, and Student-Level Factors on the Science Performance of Eighth-Grade Taiwanese Students

This study was conducted to understand the effect of student-, classroom-, and school-level factors on the science performance of 8th-grade Taiwanese students in the Trends in International Mathematics and Science Study (TIMSS) 2011 by using multilevel analysis. A total of 5,042 students from 153 classrooms of 150 schools participated in the TIMSS 2011 study, in which they were required to complete questionnaires. A 3-level multilevel analysis was used to assess the influence of factors at 3 levels on the science performance of 8th-grade Taiwanese students. The results showed that the provision of education resources at home, teachers’ level of education, and school climate were the strongest predictor of science performance at the student, classroom, and school level, respectively. It was concluded that the science performance of 8th-grade Taiwanese students is driven largely by individual factors. Classroom-level factors accounted for a smaller proportion of the total variance in science performance than did school-level factors.     

The microgenetic emergence of cooperative play in 6-year-olds developmentally at-risk

Cooperative play was investigated by a controlled pre/post-test intervention design with 28 dyads of 6-year-old students developmentally at-risk. Selection was based upon cut-off scores on a language development test and a nonverbal IQ test, and same-sex pairs were matched within classrooms. Co-variables were: socio economic status, free play time in school and pedagogical quality of the classrooms. Each child was rated with respect to temperament. Play session 1 served as pretest for both research conditions. Then five play sessions of 20 min within 3 weeks took place in the experimental condition. Eight weeks after session 6 a play session took place in both research conditions serving as post-test. After microgenetic analyses of the data, ANCOVA's were computed of time played in collaboration, and proportion of pretend play and of deep collaboration. The results revealed that the level of collaboration had improved dramatically in the experimental condition only. Moreover, the findings showed that the length of time dyads spent cooperating, and how deeply they collaborated on tasks, was related to how agreeable and emotionally stable they were. Gender effects also came forward. No long-term intervention effects were found.     

从科举历史看当今高考分层考查的可行性

历史事实证明,举荐到考试的转变是选才方式的进步,但这种选拔方式也会带来许多难以解决的矛盾,历史上就曾多次出现对于考试问题的争论与改革,过分地偏重于问题的任何一个方面都会使考试陷入困境。通过一次考试的方式难以解决考试中的诸多矛盾。我国科举历史上分层次考试的经验说明,在统一考试的基础上采取多层的考试方式,有助于维持对于考试的多重诉求间的平衡,值得我国当前的高考改革借鉴。     

隐性分层在高职英语口语教学中的应用研究

众所周知,高职英语口语教学的目标就是要培养高职学生的职业英语口语应用能力,使学生以后在工作职场中＂能用,够用＂。可是目前高职学生生源的英语基础水平参差不齐,如按传统的统一要求标准教学,最终根本无法实现这一目标。于是笔者尝试实行隐性分层教学,一学期后发现教学效果良好。     

层级鉴定:《档案法》修订后档案开放策略

本文基于新修订《档案法》对档案开放业务的相关规定和新的要求,采用文献分析法和案例分析法等研究手段,对档案开放策略这一问题进行研究。提出档案开放要采用层级鉴定的方法来实施,针对鉴定工作量大、专业人才缺乏和不同地区缺乏统筹等问题,本文提出要采取规范标准和组织分级等手段来解决。     

Bi-dimension decomposed hidden Markov models for multi-person activity recognition

We present a novel model for recognizing long-term complex activities involving multiple persons. The proposed model, named ＇decomposed hidden Markov model＇ （DHMM）, combines spatial decomposition and hierarchical abstraction to capture multi-modal, long-term dependent and multi-scale characteristics of activities. Decomposition in space and time offers conceptual advantages of compaction and clarity, and greatly reduces the size of state space as well as the number of parameters. DHMMs are efficient even when the number of persons is variable. We also introduce an efficient approximation algorithm for inference and parameter estimation. Experiments on multi-person activities and multi-modal individual activities demonstrate that DHMMs are more efficient and reliable than familiar models, such as coupled HMMs, hierarchical HMMs, and multi-observation HMMs.