基于概念树的文本自动分类系统的研究与实现

Research and Implementation of Text Categorization System Based on Concept Tree

该文分类是信息处理的重要研究方向,现在应用较多的都是基于统计的分类系统,本文介绍了一种新型的文本分类理念,通过概念符号化,使用数字化的概念而非词汇来组成特征项,能最大限度地保留文本信息,且不需要训练语料,能灵活适应不同的分类体系。接下来详细描述了领域特征信息提取的4个步骤,以及分类体系的选取与设计。最后给出了实验的测试数据,并对影响性能的一些关键因素进行了分析,指出了进一步提高分类性能的途径。

Text Categorization is an important task of information processing.The categorization systems based on statistics play a major role in practical applications currently.This paper brings forward a brand-new approach to Text Categorization.With symbolizing the concepts and not using words but numeric concepts to make up characteristic items,this system can keep the text information mostly,need no training corpus and suit different categorization architectures flexibly.It also describes the four steps about drawing domain information and how to choose and design the categorization architecture.At last it gives the evaluations and results,gives an analysis to some key factors that affect performance and point out the next approach to advance the performance of categorization.