CAJ | 학술논문

随着WEB2.0的迅猛发展，汉语情感倾向分类在许多不同的领域取得了广泛的应用。同时，文本情感倾向分类也是当前学术界的热门课题之一。本文旨在探究一种汉语情感倾向分类方法，通过构造一种自动分类系统，对商品评价信息进行正类、负类和中立的三分类。本文采用一个两级分类系统实现三分类，首先第一级将文本分为极性和中立两部分，然后第二级再将极性文本分为正类和负类。在文本分类方法方面，采用了基于情感词、基于规则和TSVM等不同的方法。本文最后组织了分类实验对系统效果加以验证，并对实验结果进行了分析。
수착WEB2.0적신맹발전，한어정감경향분류재허다불동적영역취득료엄범적응용。동시，문본정감경향분류야시당전학술계적열문과제지일。본문지재탐구일충한어정감경향분류방법，통과구조일충자동분류계통，대상품평개신식진행정류、부류화중립적삼분류。본문채용일개량급분류계통실현삼분류，수선제일급장문본분위겁성화중립량부분，연후제이급재장겁성문본분위정류화부류。재문본분류방법방면，채용료기우정감사、기우규칙화TSVM등불동적방법。본문최후조직료분류실험대계통효과가이험증，병대실험결과진행료분석。
Along with the rapid development of WEB2.0, automatic sentiment orientation classification gets globally used in various areas. Also, it is one of the most popular topics in the academia. The goal of this research is to construct an automatic classiifer and identify the sentimental orientation of merchandise comments, dividing these texts into three categories:positive, negative and neutral. This paper applies a two-level classiifcation system to accomplish the goal. First, the level-one system divides the corpus into polarity part, which consists of positive and negative texts, and neutral part. Then, the level-two system divides the polarity corpus into positive part and negative part. As for the method to texts classiifcation, different methods as sentimental word based, rule based and TSVM are used. At the end of this paper, several experiments are conducted to verify the effect of the system, followed by the analysis of the experiment results.