基于多示例学习框架的文本分类算法

针对有特殊结构的文本,传统的文本分类算法已经不能满足需求,为此提出一种基于多示例学习框架的文本分类算法。将每个文本当作一个示例包,文本中的标题和正文视为该包的两个示例;利用基于一类分类的多类分类支持向量机算法,将包映射到高维特征空间中;引入高斯核函数训练分类器,完成对无标记文本的分类预测。实验结果表明,该算法相较于传统的机器学习分类算法具有更高的分类精度,为具有特殊文本结构的文本挖掘领域研究提供了新的角度。

Text classification algorithm based on multi-instance learning framework

The traditional text categorization algorithms are unable to deal with the text with special structure.A text classification algorithm based on multi-instance learning was presented.Each text was taken as a sample package,and the title and body in the text were treated as two examples of the package.A packet was mapped into a high-dimensional feature space using a multi-class classification support vector machine algorithm based on one-class classification.Gaussian kernel function training classifier was introduced to conduct classification prediction of unmarked text.Experimental results show that the proposed algorithm has higher classification accuracy than the traditional machine learning classification algorithm,and it provides a new perspective for the research of text mining in special text structure.