从文字表记看中日外来语的异同

日语中存在大量外来语,相比之下,汉语中的外来语就相对较少。一般认为这是因为表音的片假名更加利于转写、吸收外来语,而表意的汉字比较排斥外来语。本文通过对比分析在表记外来语时汉字与片假名的特点,认为汉字的表意性虽然有其不利的一面,但也有其有利的一面,同时认为片假名也有一定的缺陷。     

连云港市金融中介支持科技企业研究

研究企业科技创新的金融支持问题,对于促进企业科技创新能力的发展,实现建立创新型国家的目标,促进经济结构转型具有极为重要的意义。选取连云港市2002—2012年间科技创新与金融支持的相关数据进行实证分析,通过单位根检验、协整分析与格兰杰因果检验得到以下结论:金融规模与科技创新长期正相关,金融效率与科技创新负相关。并在此基础上,提出了促进连云港市科技创新金融支持的意见与建议。     

A study of tailored blank welding between mild steel sheet and zn-coated steel sheet by CO2 laser beam

Research was conducted on tailored blank welding between mild steel sheet and Zn-coated steel sheet using CO₂ laser beam. The materials used in this study were low carbon steel sheets with a thickness of 1.2 mm and Zn-coated steel sheet with the same thickness and 6.3 μm Zn coating. Experiments were conducted by applying the Taguchi method to obtain optimum conditions for the application of this tailored blank laser welding method in practical manufacturing processes. Optical microscopy, XRD, SEM and TEM analysis were performed to observe the microstructures and to determine the structures of welded zone. In addition, mechanical properties were measured by the microhardness test, tensile test and Erichsen test to evaluate the formability of the welded specimen. There was no trapped Zn in the fusion zone, and the phases of this region consisted of polygonal ferrite, quasi-polygonal ferrite, banitic ferrite and martensite. The elongation value of welded specimen was more than 80% of the value in substrate metal, and the LDH value was more than 90% of the value in substrate metal.     

Banding formation and eutectic lamellar growth in directional solidified Ni50Al20Fe30 alloy

Banding formation and eutectic lamellar growth in a directionally solidified Ni₅₀Al₂₀Fe₃₀ alloy were investigated. It was found that the banding area consists of two layers. The first layer is a γ layer, while the subsequent one is a γ layer. The composition of various phases around the banding area changes with the solidification process. The banding is formed by two steps process and caused by factors such as the fractions during the sample growth process. It was found that the band was found at relatively low growth rate. Therefore. this study indicates that increasing the growth rate is an effective method to eliminate the band formation. Eutectic lamellae nucleate and grow again after the banding formation. During the initial transition lamellar growth, the relationship between the square lamellar spacing, γ₂, and the distance from, the banding, d, can be described by the following equation: γ₂ = K [1-exp(A.d)] where K and A are constant.     

在线半监督Kohonen网络的预抓取手势识别

为实现智能仿生手的抓取,提高模式识别的实时性和灵敏性,提出一种在线半监督Kohonen网络。该网络针对表面肌电信号(s EMG)的特性,在有监督Kohonen网络基础上,将有监督和无监督网络的优势进行结合,应用数据剪辑方法处理训练集更新识别网络,在线识别侧边抓取、球形抓取、三指精确抓取和圆柱形抓取4种预抓取手势。实验表明,与不同Kohonen网络相比,此识别方法具有很好的在线识别能力和正确率。     

Students' mathematics word problem‐solving achievement in a computer‐based story

The purpose of this study was to investigate the effect of a computer‐based story, which was designed in anchored instruction framework, on sixth‐grade students' mathematics word problem‐solving achievement. Problems were embedded in a story presented on a computer as computer story, and then compared with the paper‐based version of the same story and to a condition that presented the problems as typical, isolated word problems (i.e., a non‐story condition including only problems). One hundred twenty‐eight sixth‐grade students from two public middle schools in Turkey participated in this study. In a pretest–posttest experimental design, students were randomly assigned to one of the treatment groups mentioned earlier in which they solved the same mathematics word problems. A one‐way analysis of covariance was used to analyse students' achievement in the treatment groups. The results indicated that students who solved the problems in the computer story treatment had significantly higher achievement scores than students who solved the problems in the paper story and isolated word problems treatments. In addition, the story was found to be significantly more effective than the non‐story treatment when it was presented on computer.     

面向维吾尔文的多模式匹配算法研究

维吾尔文多模式匹配算法是影响维吾尔文关键词过滤和检测性能的关键步骤之一.为此,考虑维吾尔文语法特点、书写方式、字母变换形式、特殊字母等因素,提出一种基于维吾尔文音节划分的多模式匹配算法.通过Bohum-sani函数的维吾尔语音节分解方法计算字符串音节数,利用Bohum-xekli函数得到字符串音节结构,按语法特点从右至左方式进行模式比较,实现维吾尔文多模式匹配.实验结果表明,与现有模式匹配算法相比,该算法具有更高的匹配效率.     

基于分层选择策略的主动学习分词方法

为了克服训练样本不足、获取大量标注样本费时费力的问题,在基于不确定选择策略的基础上,提出了一种新的基于分层选择策略的主动学习方法。使用新提出的选择策略从大量无标注的样本中选择最有价值的样例,进行标注后加入到训练集中来训练分词器。最后在 PKU、MSR 和山西大学数据集上进行测试,并与不确定选择策略进行比较。结果表明提出的分层选择策略在相同大小的训练语料下可以获得更高的分词准确率,同时还降低了人工标注的代价。     

网络敏感信息自适应多重过滤模型研究

敏感信息过滤是既重要又复杂的任务.针对当前一些敏感信息过滤模型所存在的时间滞后、准确性低、自适应性差等问题,提出了一个敏感信息自适应多重过滤模型.该模型以互联网中文文本媒体为研究对象,采用意见挖据、机器学习、高性能计算和自然语言处理等技术,从整体和语义角度自适应识别敏感信息.对敏感信息自适应多重过滤模型的研究将为舆情监控、商业智能、辅助决策等应用系统开发提供技术支持.     

基于领域划分的微博用户影响力分析

近年来微博作为一种新兴的社交网络逐渐被广大用户使用.微博信息简短、更新迅速、包含信息量大,给微博用户获取信息带来了诸多不便,因此,利用影响力分析的手段找到具有较大影响力的微博用户具有重大意义.微博内容较传统的媒体信息具有较强的时效性和权威性,同时微博用语也极其不规范,这给微博用户影响力的分析带来了极大的困难.首先对获取的微博用户信息进行领域的划分,采用基于微博内容和用户关注的方式将用户归类到其所属的领域.其中,采用新词发现以及特征扩展的方法来提高划分结果的准确性.然后,对各个领域的用户进行影响力分析,提出3种影响力传播模型,用户最终的影响力大小根据3种模型的结果进行加权计算.最后对实验结果进行分析、比较,证明了计算用户影响力的方法能取得较优的结果.