基于潜在语义索引的特征选择与权重改进若干关键问题的研究与实现  

Research and Implementation of Several Key Problems in Feature Choice and Weight Improvement Based on Latent Semantic Indexing

作　　者：李媛媛[1] 马永强[1] 

机构地区：[1]西南交通大学信息科学与技术学院,成都610031

出　　处：《现代图书情报技术》2007年第10期80-84,共5页New Technology of Library and Information Service

摘　　要：分析潜在语义索引的基本原理及其特点。针对LSI的3个因素,从特征词的选取,维数约简,特征词权重3个方面进行约定和改进。并以计算机类的科技文献作为测试文档,对改进的权重算法和改进前后LSI系统的检索结果进行分析。结果显示,特征词的选择结果及检索效果都得到较大的提高,性能稳定。The basic theory and its features about Latent Semantic Indexing（LSI） are analyzed. For the three factors of LSI, the word selection, dimension simplification, words weighting have been engaged and improved. Scientific and technical literatures from computing are used as testing documents, also the improved weight algorithm and the retrieval results about two LSI systems are analyzed. The experimental results show that the feature choice and retrieval results are superior improved and hard performance with the new weight algorithm.

关 键 词：潜在语义 权重改进 数据稀疏 特征选择 

分 类 号：TP391.3[自动化与计算机技术—计算机应用技术]