概念语义生成与文本特征选择研究

摘要

Text feature selection is a key technology of text classification and information extraction.For text classification with high dimensional sparse feature vector problem,a feature extraction method based on non-negative matrix factorization and concept semantic space is presented.This method gives the interpretation of the theme and better reflects the local characteristics of the text by adding the non-negative limitation to the matrix factorization.Experimental results show higher accuracy of the classification is achieved in local semantic space.%文本特征选择是文本分类和信息提取的关键技术.针对文本分类中特征向量的高维稀疏问题,提出了非负矩阵分解和概念语义空间结合的特征抽取方法,对特征矩阵分解算法加入非负限制能够给出概念语义向量面向主题的解释,较好体现文本的局部特征.采用非负矩阵分解对全局和局部语义空间进行降维处理提高了体征提取效率,对不同概念语义空间中文本分类效果比对分析.实验结果表明基于非负矩阵分解的局部概念语义空间中文本分类精度较高.