Boosting text segmentation via progressive classification

A novel approach for reconciling tuples stored as free text into an existing attribute schema is proposed. The basic idea is to subject the available text to progressive classification, i.e., a multi-stage classification scheme where, at each intermediate stage, a classifier is learnt that analyzes the textual fragments not reconciled at the end of the previous steps. Classification is accomplished by an ad hoc exploitation of traditional association mining algorithms, and is supported by a data transformation scheme which takes advantage of domain-specific dictionaries/ontologies. A key feature is the capability of progressively enriching the available ontology with the results of the previous stages of classification, thus significantly improving the overall classification accuracy. An extensive experimental evaluation shows the effectiveness of our approach.     

Mining e-Learning domain concept map from academic articles

Recent researches have demonstrated the importance of concept map and its versatile applications especially in e-Learning. For example, while designing adaptive learning materials, designers need to refer to the concept map of a subject domain. Moreover, concept maps can show the whole picture and core knowledge about a subject domain. Research from literature also suggests that graphical representation of domain knowledge can reduce the problems of information overload and learning disorientation for learners. However, construction of concept maps typically relied upon domain experts in the past; it is a time consuming and high cost task. Concept maps creation for emerging new domains such as e-Learning is even more challenging due to its ongoing development nature. The aim of this paper is to construct e-Learning domain concept maps from academic articles. We adopt some relevant journal articles and conference papers in e-Learning domain as data sources, and apply text-mining techniques to automatically construct concept maps for e-Learning domain. The constructed concept maps can provide a useful reference for researchers, who are new to the e-Leaning field, to study related issues, for teachers to design adaptive learning materials, and for learners to understand the whole picture of e-Learning domain knowledge.     

The segmented presentation of visually structured texts: Effects on text comprehension

The effects of a segmented presentation applied to a visually structured text were examined in the context of the explosion of small-screen devices. Empirical research investigating the influence of text signaling on text processing suggests that the text visual structure may influence comprehension by facilitating the construction of a coherent text representation. Undergraduate students were asked to read a text under different segmented conditions varying on the type of information provided about the text visual structure and on the segmentation unit. When the segmented presentation did not supply any information or when it only offered local information about the text visual structure, text comprehension depended on the segmentation unit. When the segmentation unit did not fit the text visual structure, an erroneous text representation was constructed, whereas the compatible segmentation unit led to a correct text comprehension. When the segmented presentation rendered the global text visual structure, the segmentation unit had no effect on comprehension and more readers constructed a correct and close text representation. Thus, the text visual structure seems to play a role in text comprehension and this role has to be taken into account for text segmented presentation.     

基于文档标引图模型的文本相似度策略

文档标引图是一种基于短语的图结构文本特征表示模型,能更加全面、准确地表达文本特征信息,实现渐增的文本聚类和信息处理。该文基于文档标引图特征模型,提出文档相似度计算加法策略和乘法策略,采用变换函数对文档相似度值进行调整,增强文档之间的可区分性,改进文本聚类和分类等处理的性能,实例证明了策略的有效性。     

基于文字下划线特征变化的脆弱型文本水印

根据特征编码的思想,首次提出了一种基于文字下划线特征变化的脆弱型文本水印算法.该算法利用字符的内码特征信息生成水印,然后按照特定的顺序插入水印或检测水印.理论分析和实验表明,算法具有良好的不可感知性,能准确地检测各类水印攻击,在大部分情况下能对篡改进行精确定位.     

一种结合结构化信息的富文本分类方法

随着文本表现形式越来越丰富,文本分类研究的对象正从平文本逐渐转变为富文本,传统的平文本分类方法不能满足实际需要.分析了富文本中的结构化信息和文本内容信息,把它们作为两个重要的因素,综合考虑了其在分类中的作用,提出并实现了标签组件法、结构组件法和综合法三种富文本分类的方法.实验表明,所提出的方法有较好的分类表现,能解决OpenDocument的分类问题.     

DTD约束下的XML树模式查询最小化

目前大部分XML查询语言都使用树模式来匹配待查询的XML文档树以得到所需要的、与模式树相吻合的查询结果,此效率在很大程度上取决于XML模式树的大小,那么尽可能快速地查找并删除查询模式树中的冗余节点就变得十分重要。重点讨论DTD约束下树模式的最小化问题,将DTD兄弟约束SC拓展成扩展兄弟约束ESC,使其能够表达DTD约束中的祖先-后代关系;并指出只包含{ESC,/,//,[],*}的查询树模式的最小化问题的复杂度是指数级的,且当模式树是分支受限的时候,其最小化问题的复杂度是多项式时间的;最后给出了一个多项式时间的受限分支的模式树最小化算法。     

基于矩阵谱分析的文本聚类集成算法

聚类集成技术可有效提高单聚类算法的精度和稳定性,其中的关键问题是如何根据不同的聚类成员组合为更好的聚类结果.文中引入谱聚类算法解决文本聚类集成问题,设计基于正则化拉普拉斯矩阵的谱算法(NLM-SA).该算法基于代数变换,通过求解小规模矩阵的特征值和特征向量间接获得正则化拉普拉斯矩阵的特征向量,并用于后续聚类.进一步研究谱聚类算法的关键思想,设计基于超边转移概率矩阵的谱算法(HTMSA).该算法通过求解超边的低维嵌入间接获得文本的低维嵌入,并用于后续K均值算法.在TREC和Reuters文本集上的实验结果验证NLMSA和HTMSA的有效性,它们都获得比其它基于图划分的集成算法更为优越的结果.HTMSA获得的结果比NLMSA略差,而时间和空间需求则比NLMSA低得多.     

一种基于DOM的Web文档主题划分方法

主题划分是多主题文档自动摘要中的一个重要问题,提出了一种以网页结构为指导,利用页面对应DOM树中节点的自然分割功能以及相邻边界节点语义相似度的比较进行网页主题划分的方法.实验结果表明该方法具有较高的划分准确率,在此基础上抽取的网页摘要可显著增加文摘内容对原文的覆盖率、有效解决Web文档摘要分布不平衡问题.