Emotions elicited while drinking coffee: A cross-cultural comparison between Korean and Chinese consumers

Consumers assess products not only based on their physical aspects but also on their emotion-evoking components. Many methods, such as EsSense Profile^®, GEOS, and ScentMove, have been developed for evaluating various kinds of products. Research has suggested the need for developing a specialized lexicon for specific food or beverage categories. The English emotion lexicon, the Coffee Drinking Experience¹(CDE), was developed specifically for coffee. However, while emotions can be expressed by words, they can also be affected by culture and context. Thus, our objectives are 1) to develop coffee emotion lexicons in Korean and Chinese, 2) to compare the differences of emotions evoked while consuming different kinds of coffee, and 3) to compare the differences between Chinese and Korean consumers while consuming the same coffee samples. To address these aims, we conducted two studies. First, we conducted a consumer-led translation of 86 emotion terms from the CDE into Korean and Chinese. Second, consumers checked the intensity of emotion terms while drinking six commercial coffee samples. The order of the Chinese and Korean consumers’ overall preferences was the same: all-in-one coffee received higher scores than black coffee. Finally, using a logistic stepwise regression analysis, the Chinese and Korean lexicons were reduced to 53 and 29 terms, respectively.     

一种基于文本分类和评分机制的软件缺陷分配方法

开源软件项目的缺陷管理和修复是保障软件质量及软件开发效率的重要手段,而提高软件缺陷分配的效率是其中亟需解决的一个关键问题。文中提出了一种基于文本分类和评分机制的开发者预测方法,其核心思想是综合考虑基于机器学习的文本分类和基于软件缺陷从属特征的评分机制来构建预测模型。针对大型开源软件项目Eclipse和Mozilla的十万级已修复软件缺陷的实验表明,在“十折”增量验证模式下,所提方法的最好平均准确率分别达到了78.39%和64.94%,比基准方法(机器学习分类+再分配图)的最高平均准确率分别提升了17.34%和10.82%,从而验证了其有效性。     

基于图的混合推荐算法*

混合推荐是解决各种单一推荐方法缺陷的重要途径,文中提出基于图的混合推荐算法,通过在图中融合各种推荐因素进行建模,产生最终的推荐结果.利用推荐物品的内容属性计算物品间的相似度,构建最近邻图关联矩阵.根据物品的打分记录构建物品的兴趣模型,生成矢量函数.在此基础上,利用正则化框架组合关联矩阵和矢量函数,构建基于图的学习模型,实现基于图的混合推荐,并从理论上证明算法的收敛性.在MovieLens数据集和亚马逊网上商城交易数据上的对比实验验证文中算法的有效性.     

Estimation of irradiance and outdoor performance of photovoltaic modules by meteorological data

Three environmental factors of irradiance, solar spectral distribution and module temperature greatly affect the performances of photovoltaic (PV) modules. If the environmental factors can be estimated by basic meteorological data (BMD) announced by an official organization in various areas, the performances of PV modules can be estimated easily. In this study, a relationship between the environmental factors and the BMD was analyzed. The performances of Si-based (crystalline Si and thin-film Si) PV modules were estimated by the relationship. As a result, errors between the estimation and actual performances in the crystalline Si and thin-film PV modules were within 1.88% and 0.58% in Kusatsu city, Japan (34°58′N, 135°57′E). This methodology can be useful for rating the performance of PV modules at various areas.     

Executive function in individuals with subthreshold autism traits.

Objective: Recent research has documented increased psychosocial difficulties in individuals who report higher-than-typical autistic traits but without an Autism Spectrum Disorder (ASD) diagnosis. Less is known, however, regarding the cognitive profile of individuals with subthreshold autism symptomatology. The objective of the present study was to provide additional insight into this issue and examine whether young adults who report higher degrees of autism traits also report experiencing increased difficulties with executive control. Method: The Behavior Rating Inventory of Executive Function was utilized to evaluate behavioral aspects of executive functioning in 66 and 28 individuals who endorsed high and low subthreshold levels of autism symptomatology, respectively. Results: After accounting for Attention Deficit/Hyperactivity Disorder (ADHD) symptomatology at both the group and individual participant levels, we found that autism traits continued to explain a significant amount of variance in participants' overall level of executive function (Global Executive Composite) as well as within most individual executive domains. Interestingly, the high and low trait groups did not differ on the inhibitory control and organization of materials scales, areas of functioning that appears to be largely spared in individuals with ASD as well. Conclusions: Findings from the present study are consistent with past research linking ASD and executive control impairment. In addition, ASD and ADHD traits were associated with unique contributions to the executive control profile of individuals with subthreshold autism symptomatology. This finding underscores the importance of accounting for ADHD symptomatology in studying ASD. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

面料评级方法在提高针织纱质量中的应用

为提高Modal针织纱质量,运用面料评级与乌斯特检测相结合的方法进行纺纱质量监控,可准确、快捷地找出质量问题产生的原因,并及时采取纠正措施,取得了良好的效果。认为,面料评级方法更加接近市场用户的质量标准,由于抽样数量较多,代表性较强,便于对布面疵点进行直观分析,有利于准确查找原因,并有针对性地进行质量改进,可有效地避免质量问题给用户带来的损失;与乌斯特检测相结合,能够取得良好的质量控制效果。     

数字电视管理系统的研究与实现

提出了基于IPv6组播技术与G.1070算法的数字电视管理系统;通过捕捉并分析网络数据包,对ITU-T推荐的G.1070算法进行改进后,得出各收视终端的QoE量化指标;通过分析MLDv2协议的消息类型得出各收视终端的收视率。作为中间件,该系统可以无缝地嵌入到各厂商生产的网络收视终端。     

管材矫直机减速器箱体性能评价的研究

矫直机减速器是应用于金属型材、棒材、管材和线材等矫直设备中的关键部件之一,工作过程中需承受较大的冲击载荷和过载,故其性能的好坏直接影响整个矫直设备的可靠性。针对某款管材矫直机用减速器工况复杂恶劣的特点,采用有限元法对矫直机减速器箱体进行强度计算和固有特性分析,通过计算箱体的应力应变及固有频率,研究主机及传动系统对减速器箱体的影响,从而对箱体进行性能评价,保证箱体的可靠性和齿轮箱系统运行的平稳性,最后通过实践应用验证了评判方法的正确性。     

基于Widrow-Hoff神经网络的多指标推荐算法

为解决传统的协同过滤推荐算法不能综合运用多个指标进行推荐的问题,通过引入多指标评分的概念对标准的协同过滤推荐算法进行扩展,提出一种基于Widrow-Hoff神经网络的多指标推荐算法。利用Widrow-Hoff最小二乘法自适应算法在进行系统辨识时的高精度拟合特性,提出一种基于Widrow-Hoff最小二乘法算法的用户偏好特征向量计算方法。利用用户偏好特征向量和空间距离矩阵度量用户相似度,以定位邻居集并为用户推荐最优项目。实验结果表明,本文算法可提高推荐精度,改进推荐质量。     

融合信息熵与信任机制的防攻击推荐算法研究

由于对用户偏好信息的过分依赖,致使推荐系统易受到恶意攻击,从而影响系统的推荐质量。提出一个融合信息熵与信任机制的防攻击推荐算法。在考虑了托攻击与正常用户之间的评分变化幅度差异基础上,提出融合信息熵的相似性改进算法,同时引入信任更新机制,在推荐过程中将用户间信任度与相似度有机相结合,通过筛选推荐权重较高的邻居用户方法获得可靠推荐,从而降低恶意攻击对系统的影响。通过在真实数据集上实验表明该算法在提高推荐系统的准确性和脆弱性上有较好的表现。