层次非负矩阵分解及在文本聚类中的应用

文本聚类的目标是把数据集中内容相似的文档归为一类,而使内容不同的文档分开。目前针对不同领域的需求,多种解决聚类问题的算法应运而生。然而,由于文本数据本身固有的复杂特点,如海量、高维、稀疏等,使得对海量文本数据的聚类仍然是一个棘手的问题。提出了层次非负矩阵分解聚类方法,该方法不但保留了非负矩阵分解的优点,如同步识别文档类别和找出类别本质特征,而且能够展现类别间的层次结构。这种类别层次结构在网页预览等应用中是非常有用的。在真实数据集20Newsgroups和Reuters-RCV1上的实验结果表明,层次非负矩阵分解相比已有的方法更有效。     

一种改进的谱聚类方法及其在文本分析中的应用

随着文本资源的激增,特别是网页文本的迅速增加,针对文本的挖掘分析日益受到重视。谱聚类是文本聚类分析较常用的一种新型方法。该文将非负约束引入到传统的谱聚类算法中,提出了一种基于非负约束的谱聚类方法。文中实验验证了所提出方法在中文文本聚类分析应用中的有效性。     

基于RBF神经网络的Web文本分类的研究

Web文本分类是采用文本分类技术将Web上的信息进行自动分类,使用户能够快速找到自己想要的资源。文本分类的过程中,将特征提取之后的来自Web的数据分成样本数据集和测试数据集,将样本数据集输入到RBF网络中进行训练,RBF网络经过训练之后,输入测试数据集中的数据进行验证,实验证明,RBF网络取得了较好的分类结果。     

留白之美——版式设计中空白的运用

所谓版式设计是指在有限的版面空间里,将版面的构成要素——点、线、面,根据特定内容的需要进行组合排列,并运用造型要素以及形式原理,把构思与计划以视觉形式表达出来,寻求艺术手段来正确地表现版面的信息。在版式设计中一般设计师都会把重点放在黑的设计上,却容易忽视白的设计。文章以版式设计教学为例,阐述版式设计中白的作用及其地位,并探讨版式设计中如何恰当使用空白。     

Organizational Norm Congruency and Interpersonal Familiarity in E‐Mail: Examining Messages From Two Different Status Perspectives

Relying on SIDE theory, this 2 × 2 experimental design tested the effects of adherence to organizational norms and interpersonal familiarity in e‐mail messages from both superior and subordinate perspectives. Results reveal that using norm‐congruent e‐mail messages account for over 50% of the variance in both superior and subordinate attitudes toward the message and between 30% and 56% of the variance in perceptions of source credibility. Data from the superiors indicate that norm congruent e‐mail messages account for 14.4% of the variance in message compliance. Results from subordinates reveal an interaction effect that is consistent with SIDE‐based predictions. Subordinates believe superiors are most likely to comply with e‐mail requests from unfamiliar subordinates that are crafted in a manner congruent with organizational norms.     

‘There was a lot of learning going on’ Using a digital medium to support learning in a professional course for new HE lecturers

This small scale action research study investigated the experiences of learners over two iterations as they completed a patchwork text assignment within the digital medium of a personal learning system (PLS). The aim was to investigate the extent to which using a PLS can facilitate formative and collaborative feedback to assist student learning. A secondary question, linked to the course being a post graduate Certificate for new HE teachers within a geographically dispersed university, was to examine if using a PLS in the programme would influence their interest in using it with their own students in the future. This question is of particular importance for the tutors of this programme, as the university attracts mainly part-time students who are studying at a distance from the university campus. The study was carried out as part of the JISC funded Flourish Project, whose overall aim was to examine the use of a PLS to support the development of academic staff.     

向量空间模型中结合句法的文本表示研究

为增强向量空间模型(VSM)中项的语义描述性,克服VSM中各语义单元相互独立的缺陷,提出一种基于短语的特征粒度描述方法。该方法从文本的表示及特征项之间的组织方式入手,通过句法规则识别基本短语,构建特征与中心动词的关系树,利用基本短语代替BOW中的词。实验结果表明,采用基本短语的文本表示可提高分类的性能,增加项之间的联系,克服特征项相互独立的缺陷,在特征数量较少的情况下仍能保持良好的分类效果。     

基于后验概率制导的B-KNN文本分类方法

针对K最近邻(KNN)方法分类准确率高但分类效率较低的特点,提出基于后验概率制导的贝叶斯K最近邻(B-KNN)方法。利用测试文本的后验概率信息对训练集多路静态搜索树进行剪枝,在被压缩的候选类型空间内查找样本的K个最近邻,从而在保证分类准确率的同时提高KNN方法的效率。实验结果表明,与KNN相比,B-KNN的性能有较大提升,更适用于具有较深层次类型空间的文本分类应用。     

信息检索中相关实体发现综述

实体是Web页面中的重要信息载体,用户通过搜索引擎进行信息检索中时一般想得到某个具体的实体,而不是某些文档的列表,因而信息检索中的相关实体发现研究就具有非常重要的意义。对信息检索中的相关实体发现的基本过程进行了综述,重点描述了相关实体发现的重要组成部分:全文检索、实体识别、实体分级,主页查找及其各部分所涉及到的关键问题。     

Using enhanced text to facilitate recognition of drug names: evidence from two experimental studies

Two studies examined the effects of enhanced text on immediate recognition of drug names. Study 1 sampled 102 college students using a between-subjects design, while Study 2 sampled 11 practicing pharmacists and technicians using a within-subjects Latin square design. Both studies utilized a computer-based sequential recognition task where a prime word was shown with various text enhancements, followed by a mask and then a target word. Participants decided whether the target word was the same as the prime word. Stimuli were organized so that 120 trials were matches and 120 trials were mismatches, randomly sequenced for each participant. Results showed an effect of orthographic similarity, where high-similarity name mismatches were missed more often. This effect was independent of the type of text enhancement used and response bias. Case-based enhancements also increased errors of commission (false alarms) significantly. Discussion includes the practical relevance of the data and future directions for research.