A hybrid generative/discriminative approach to text classification with additional information

This paper presents a classifier for text data samples consisting of main text and additional components, such as Web pages and technical papers. We focus on multiclass and single-labeled text classification problems and design the classifier based on a hybrid composed of probabilistic generative and discriminative approaches. Our formulation considers individual component generative models and constructs the classifier by combining these trained models based on the maximum entropy principle. We use naive Bayes models as the component generative models for the main text and additional components such as titles, links, and authors, so that we can apply our formulation to document and Web page classification problems. Our experimental results for four test collections confirmed that our hybrid approach effectively combined main text and additional components and thus improved classification performance.     

The Persistence of Solid and Liquid Naive Conceptions: A Reaction Time Study

The study explores whether the naive concepts of solid and liquid persist in adolescence. Accuracy of responses and reaction times where measured while 41 ninth graders classified different solids (rigid, non-rigid and powders) and different liquids (runny, dense) into solid or liquid. The results show that these naive conceptions affect adolescences’ classifications in terms of both accuracy and reaction time. The rate of correct classifications of non-rigid solids and powders was significantly lower than of rigid solids. Lower rate of success was also found for classification of dense liquids compared with runny liquids. In addition, the reaction time results of correct classifications for non-rigid solids and powders were longer than those for rigid solids and, likewise, reaction times for dense liquids were longer than for runny ones. These results suggest that reasoning processes associated with correct classification of objects that are not consistent with the naive conceptions are more demanding.     

混合多层分类和朴素贝叶斯模型的垂直搜索引擎分类器设计

研究Web上计算机教育资源的分布特点,融合主题词和文档形式,设计多层分类器来完成主题搜索过程中的正确分类,继而应用朴素贝叶斯模型对主题资源信息进行自动类别划分,完成资源的物理存储。实验中主题分类的平均正确率约78%,主题的平均召回率约61%,而资源解析的平均正确率约81.5%,测试结果能够验证本文设计思想的可行性。     

帮助儿童学会反思——来自心理理论研究的启示

作为朴素理论的一个核心领域,儿童心理理论研究是目前发展心理学研究中的前沿课题之一.研究表明,儿童不仅能发现人是有心智的,是由各种心理状态构成的,而且能发现可以通过影响一个人的想法而改变他的行为.教育者应该把儿童看成是拥有各种心理状态的、会思考的个体,在教育过程中提供相应的教育策略,帮助儿童成长为真正意义上的能思考、会反思的人.     

基于雨林框架的朴素贝叶斯分类算法

出了一种可伸缩的朴素贝叶斯分类算法。算法针对大数据集的训练数据,通过构建雨林框架,能在有限主存里存储训练数据,训练生成概率矩阵,进而对测试样本进行分类。算法仅对整库一次扫描。实验表明,该算法能够获得与整库读入主存相同的分类准确率．并且有较高的处理效率。     

基于本体和加权朴素贝叶斯的网络舆情主题分类

及时准确地对舆情信息进行主题分类,不仅能实时了解舆情动态变化,还能为预判舆情发展趋势、舆论引导建立基础。本文提出一种基于本体和加权朴素贝叶斯的网络舆情主题分类方法,通过使用本体将领域知识和领域文本特征融入分类过程中。将该方法应用到动物卫生领域舆情主题分类中,分类结果精确度为0.9402,Marco_F1达到0.9339。通过与朴素贝叶斯（NB）和THUCTC两种方法的对比实验,证明本文提出的基于本体和加权朴素贝叶斯的分类方法有效且具有可行性,但是领域本体的概念、关系的完备程度会影响分类的效率。     

改进朴素贝叶斯算法的人脸表情识别

传统图像特征提取具有较高维度缺陷,造成算法分类效率低、复杂度高、分类速度慢、计算开销大等问题.为此提出AAM算法,定位关键点提取人脸表情几何特征.将朴素贝叶斯分类器结合特征属性重要度调节高斯核函数,使用K近邻算法实现分类决策,提出一种WNBC-KNN分类方法,从降低数据维度和分类算法两方面优化人脸表情分类.在CK+数据...     

垃圾邮件的过滤技术研究

简单阐述了垃圾邮件的定义、现状和危害，对目前垃圾邮件的主要几种过滤技术进行了全面介绍和分析，并指出了垃圾邮件过滤技术的发展趋势。     

一种针对已知作者的姓名消歧方法

在外文期刊数据库中,同一姓名简称代表多位作者的现象十分普遍,严重影响作者检索的精度。本次研究将规则与算法相结合,依据规则为分类算法标注训练数据,从而在无监督条件下使用有监督算法,实现作者的精确检索。该算法适用于论文查证等已知作者身份的姓名消歧问题,相比通用的消歧方法,该方法结合无监督算法无需人工标注的优点,以及有监督算法高效率、易对应实体的优点。实践结果表明,该方法具有较高的准确度。