摘要
以科技文献中的共现关键词—叙词词对(以下简称共现词对)为研究对象,运用互信息法、概率法计算方法,以准确率和召回率为衡量指标,通过绝对值方式和相对最高值方式,分析这两种统计方法在不同区间内对同义共现词对的自动识别效果及特点。研究发现,概率法的相对最高值方式具有最佳的同义共现词对识别效果。最后,探讨互信息法、概率法与共现频率的关系,揭示这两种统计方法在同义共现词对的识别上存在互补的特性。
Mutual information and probability method, which statistic results classified into absolute value and relative maximum value, will he used to makes a comparative analysis of the automatic recognition of synonymous on co-occurrence keywords and descriptor.Taking the precision and recall rate as measurement index, the relative maximum value of probability method can get a better synonymous recognition performance in the co occurrence keywords and descriptor. At last, the paper discovers complementary functions in the mutual information and the probability method by investigated the relation of the frequency of the co-occurrence between keywords and descriptor and the two statistic method.
出处
《图书情报工作》
CSSCI
北大核心
2012年第18期122-126,共5页
Library and Information Service
基金
教育部人文社会科学研究一般项目“共现词汇语义关系挖掘与本体自动构建研究”(项目编号:10YJC870051)
广东高校优秀青年创新人才培育项目“基于词共现的关键词语关系自动识别研究”(项目编号:wym09089)研究成果之一
关键词
同义词识别
互信息法
概率法
共现词
synonyms recognition
mutual information method
probability method
co-occurrence word
