摘要
提出带权约简的概念,并研究了带权约简算法.首先指出已有约简算法无法融合人类的先验知识;然后提出使用权值向量表示这类知识,用于属性重要性的计算,获得基于区分能力的带权约简算法,并分析带权约简与经典约简的关系;最后将算法应用于汉语词性标注自动校对,并讨论了权值向量的具体设置.实验结果表明,使用所提出的算法及相应权值向量,可获得更有利于预测的约简.
The concept of weighted reduct is introduced and a weighted reduction algorithm is proposed, in which the weight vector represents knowledge of human experts. The algorithm is an extension of the reduction algorithm based on discernibility, and weighted reduets are more general than traditional reduets. The algorithm is applied to automatic correction of Chinese part-of-speech tagging, and experimental results show that reducts with better prediction potential are obtained by using appropriate setting of the weight vector.
出处
《控制与决策》
EI
CSCD
北大核心
2007年第7期740-744,共5页
Control and Decision
关键词
知识约简
带权约简
属性重要性
汉语词性标注
自动校对
Knowledge reduetion
Weighted reduetion
Significance of the attribute
Chinese part-of-speech tagging
Automatic correction
