摘要： 提出了一种与分词一体化的人名识别方法，根据中文人名内部用字产生潜在人名，可信度较高的潜在人名与其它候选切分词共同组成分词有向图的节点。利用Bigram和Trigram给有向边赋值，使有向图的最短路径对应句子的正确切分，确定了句子的切分路径即可识别出句子中的人名。实验结果表明，该方法取得了较好的人名识别正确率。

关键词: 中文人名识别, 未登录词识别, 汉语自动分词, 最短路径

Abstract: This paper presents a method of Chinese personal name recognition synchronized with Chinese word segmentation. Possible person names are generated according to their internal characters and those with high confidence are added into the segmentation digraph as vertices along with lexical word candidates. Bigram and Trigram are used to evaluate each path of segmentation digraph, thus the shortest path corresponds to the optimal segmentation of the Chinese sentence. Personal names can be recognized once the optimal segmentation is selected. The experimental results show the method achieves satisfying accuracy.

Key words: Chinese personal name recognition, Unknown words recognition, Chinese word segmentation, Shortest path

中图分类号: 

• TP391.1