摘要
观点挖掘近年来已经成为自然语言处理领域的热点问题,该文对观点挖掘的几项关键技术—评价对象、评价短语、主观性关系抽取、倾向性判断进行了研究。在评价对象抽取阶段,通过统计得到所有的名词和名词短语作为候选,然后结合词频,词共现等特征进行过滤得到最终的评价对象;在评价短语抽取阶段,使用基于观点词词典的匹配方法,并把观点词前面的副词也作为评价短语的一部分;在搭配关系抽取阶段,目的是抽取评价对象和评价短语的关联关系,采取的方法是将在句中距离评级对象最近的评价短语作为该短语的评级短语;在情感倾向分析阶段,通过将情感句进行分类,然后制定规则进行无监督的倾向性判断。
In recent years,opinion mining has become a hot issue in natural language processing field.This paper carries out a re search on several key technologies of opinion mining,such as the object of evaluation,evaluation phrases,subjective relationship extraction and orientation judgment.In the phase of evaluating objects,all the nouns and noun phrases collected through statistics are regarded as candidates and then we associate them with word frequency and co-occurrence to filtrate and finally gets the ob ject of evaluation.In the phase of evaluation phrases extraction,the writher applies a matching methodology which is based on af fection and takes the adverbs which precede the affective words as a part of evaluating phrases.The goal in the phase of subjective relationship extraction is to extract the incidence relation between evaluation object and evaluation phrases and the solution that we have adopted is to take the evaluation phrase which is close to the evaluation object as rating phrases.In the phase of orienta tion judgment,we classify the affective sentences and then develop rules for non-supervised orientation judgment.
出处
《电脑知识与技术》
2013年第3X期2044-2045,2047,共3页
Computer Knowledge and Technology
关键词
观点挖掘
评价对象
评价短语
主观性关系
倾向性判断
opninion mining
target
evaluation phrase
subjective relationship
orientation judgment
