基于余弦相似度的文本空间索引方法研究

An Approach for Spatial Index of Text Information Based on Cosine Similarity

基于相似度的数据空间索引在数据挖掘及数据可视化等方面有着重要的应用。本文以新闻的标题为研究对象,提出了以 CrossAVL为基础的文本对象层次式聚类方法以及文本信息空间索引算法 FastMap-MDS,有效地保持了文本对象间的相似信息。实验表明,该方法具有较高的效率和精度。

Spatial index for data based on similarity can be employed by applications on data mining and data visualization widely. To build spatial index of news title, this paper implements hierarchical cluster algorithm for news titles with CrossAVL as data structure for the similarity matrix storing and presents an available and efficiency method named as FastMap-MDS. Experiment results show that this method can work efficiently while the similarity information are kept well.