Efficient SVM-based recognition of Chinese personal names

This paper provides a flexible and efficient method to identify Chinese personal names based on Support Vector Machines (SVM). In its approach, forming rules of personal name are employed to select candidate set, then SVM based identification strategies are used to recognize real personal name in the candidate set. Basic semanteme of word in context and frequency information of word inside candidate are selected as features in its methodology, which reduces the feature space scale dramatically and calculates more efficiently. Results of open testing achieve F-measure 90.59% in 2 million words news and F-measure 86.67% in 16.17 million words news based on this project.