文本挖掘技术研究进展

文本挖掘是一个对具有丰富语义的文本进行分析从而理解其所包含的内容和意义的过程,已经成为数据挖掘中一个日益流行而重要的研究领域。首先给出了文本挖掘的定义和框架,对文本挖掘中预处理、文本摘要、文本分类、聚类、关联分析及可视化技术进行了详尽的分析,并归纳了最新的研究进展。最后指出了文本挖掘在知识发现中的重要意义,展望了文本挖掘在信息技术中的发展前景。     

Natural language understanding within a cognitive semantics framework

The article argues that cognitive linguistic theory may prove an alternative to the Montague paradigm for designing natural language understanding systems. Within this framework it describes a system which models language understanding as a dialogical process between user and computer. The system operates with natural language texts as input and represent language meaning as entity-relationship diagrams.     

基于支持向量机的中文文本自动分类研究

首先对文本提取特征向量,再利用词语相似度求出文本特征子集,由支持向量机进行文本分类,实现了一个中文文本自动分类系统,并对该系统进行了针对SVM大规模真实文本的试验测试。试验表明,该方法的系统的招回率较低,而准确率较高,取得了令人满意的结果。     

基于灵长类视觉皮层的目标识别模型综述

在过去的十年,通过实验手段,已积累灵长类视觉皮层的大量实验数据.许多研究者根据这些实验数据提出动物视觉系统的模型,并将它们用于计算机目标识别.文中首先总结动物视觉系统的主要实验结论,然后评述基于灵长类视觉系统工作机制的常用模型,重点讨论MIT的皮层等级模型.     

From perceptual to conceptual categorization of wines: What is the effect of expertise?

Wine supply in the French market is structured in an intricate system of categories based on origin. There is very little knowledge about consumers understanding of this complex category system and the sensory styles behind these categories. This study investigated how assessors with different level of expertise categorized Beaujolais wines from general to more specific levels of categorization (grape variety, appellation, and “lieu-dit”) in both perceptual (wines) and conceptual (wine labels) conditions. Based on the literature on expertise, we expected a stronger effect in the perceptual condition, in particular for the most specific levels of categorization. For each wine categorization level, three sets of 12 wines were tasted by three panels of 60 assessors: a panel of unfamiliar novices with no much exposure to Beaujolais wines; a panel of familiar novices with regular exposure to Beaujolais wines; and a panel of experts from the Beaujolais. In both perceptual and conceptual conditions, assessors were asked to perform a binary sorting task, followed by a verbalisation task. Data were analysed using DISTATIS. At each level, with a few exceptions, a clearer separation was observed between the two categories in the conceptual condition than in the perceptual condition. Although the experts categorized the labels by grape variety, they did not spontaneously categorize the wines in that way. Finally, we observed a clear expertise effect only in the conceptual condition. This result will be discussed in terms of expertise acquisition and categorization processes.     

一种K-means聚类和超球结合的多类分类算法

提出一种k-means聚类算法和超球结合的多类分类算法。对每一类样本,使用k-means算法获得子类,再在各个子类上构造最小超球,由此,对每类都获得一个超球集,这些超球将样本空间分割,根据样本点所在空间的位置,综合得到决策函数,用于对输入样本点进行类别判断。     

一种改进的SVM文本分类算法

在研究基于支持向量机进行文本分类一般步骤的基础上,针对Key-Substring-Group文本分类算法存在非线性支持向量机对核函数和参数C的强依赖的问题,用欧氏距离代替支持向量机训练得到的分类决策面进行分类决策,对文本分类算法进行改进。通过对比试验,发现分类效果不会随着核函数及参数C的变化而有明显的波动。     

Extrinsic attributes responsible for red wine quality perception: A cross-cultural study between France and Spain

Understanding quality perception requires focusing both on product proprieties and consumers’ representations. The goal of this study was to access the representations of perceived extrinsic wine quality of consumers from two different wine regions in two European countries with an important wine tradition: La Rioja-RJ-(Spain) and Côtes du Rhône-CdR-(France).Twenty commercial red wines (ten from RJ and ten from CdR) were evaluated in terms of quality by means of a categorization task. Ninety-five French and 93 Spanish consumers had to categorize the 20 wines according to four levels of quality going from very low to very high quality.Three clusters of French and two of Spanish consumers have been identified. Country-of-origin of wine arises as the most important extrinsic cue for less-involved French and Spanish consumers, while more-involved consumers with higher knowledge in wine are able to interpret and use a wider range of cues. Consumers’ knowledge, involvement and nationality appear to be good predictors of quality perception.These results are relevant for the field of sensory analysis in that they help understanding the construction of quality perception. The wine industry can benefit also from these results, especially when exporting wines. This cross-cultural study may help them to develop marketing strategies able to enhance the quality perception of wines and to make use of different extrinsic factors.     

中英文文本分类系统异同因素的探讨

论文介绍了文本分类技术的研究进展及发展方向.给出了中英文文本分类系统构建的一般模型和过程,分析总结了中英文文本分类过程中的异同,并介绍了针对这些异同因素常用的处理方法.