自然语言理解心理学在短文本分类中的实证研究  

Research of Natural Language Understanding of Psychology in the Short Text Classification

作　　者：盛宇[1] 刘俊熙[1] 郭金兰[1] 龙怡[1] 

机构地区：[1]上海政法学院计算机教研室,上海201701

出　　处：《现代情报》2009年第8期4-7,共4页Journal of Modern Information

基　　金：上海政法学院计算机实验室决策支持系统项目

摘　　要：目前对文本分类研究多数集中在对大规模语料基础上的特征选择或分类器算法的研究。本文是建立在训练样本少且样本长度短的基础上,根据人脑对自然语言理解的心理学原理"人们总是根据已知的最熟悉的、最典型的例子进行判断,只有在该方法不奏效的时候才使用频率这一概念,并且使用的是十分简单的频率"从该角度进行短文本分类的实证研究。以心理学中的"熟悉原理"、"典型原理"等为模型建立特殊词库和典型案例词库,改进了传统文本分类的实验步骤,同时提出了该方法的优势和局限性。The current research of classification of most text focused on large - scale corpus on the basis of choice of the characteristics or classification algorithm. Tiffs article is built on less training samples with short length, according to the human brain＇s understanding of the psychology principle of natural language ＂People always make a judgment according to the most familiar, the most typical example. They only use the concept of frequency when this method is not effective. It is also a very simple freguency.＂ We do research from this perspective of the short text classification. We establish a special vocabulary and a typical vocabulary based on ＂familiar principle＂, ＂typical principle＂ which are known in psychology. We improve the experimental steps of the traditional classification of the text and mention the advantages and limitations of this method.

关 键 词：文本分类 短文本 特征选择 自然语言 心理学 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]