According to the possible expressions as the features in the task, the application of Part of Speech(PoS) used in the task of Chinese personal name recognition is discussed based on the Conditional Random Fields(CRFs) on the character level. And the method of combined PoS and word-edge as a feature item is put forward. By sequence labeling on common corpus, multiple experiments of Chinese personal name recognition are token which are done in similar experiment environment with multiple applications of PoS features. Through the results of the experiments, the combination of second level PoS and word-edges is obtained the best effect in the system performance and the recognition of Chinese named entities.%根据词性在任务中可能出现的特征表达,在字粒度一级,基于条件随机场模型,对词性特征在中文命名实体识别任务中的应用进行研究,提出一种将词性和词边界合成为一个特征项的方法.在相同实验环境下针对多种词性特征的应用情况,采用序列标注的方式在公共语料上进行多次中文命名实体识别实验.通过对多次实验结果的比较分析得出,二级词性与词边界合成的特征在系统执行性能和识别效果等方面均为最优.
展开▼
