论文类型：会议论文
收录刊物：CPCI-S
页面范围：73-78
关键字：word sense disambiguation; Bayes; bootstrapping; grouping strategy
摘要：In this paper, a bootstrapping algorithm with grouping strategy is proposed for the task of Chinese word sense disambiguation. Chosen Bayes classifier as basic classifier, the classifier is updated by training the original classifier with the labeled example set which is expanded each in turn by throwing new seeds into it. In order to get the most confident seeds, the newly labeled examples are grouped according to the numbers of their features. As shown in the experimental results, the method proposed in this research overperforms basic Bayes classifier without changing other experiment conditions.