Hierarchical Conditional Random Fields (HCRF) for Chinese Named Entity Tagging

Named entity tagging is one of the key techniques in natural language processing tasks such as information extraction, answer question and so on. We present a method of Chinese NE tagging using hierarchical conditional random fields. This study is concentrated on person names, location names and organization names. We divide the process of Chinese NE tagging into three layers: person CRFs layer, location CRFs layer and organization CRFs layer. The method is characterized as follows: firstly, rich features are utilized by this model in order to increase the good performance; secondly, the hierarchical property satisfies the characteristics of Chinese NE. The experiment shows that the HCRFs model could achieve preferable results of Chinese NE tagging, in which the F value achieves 95.44%, 93.13% and 87.14 for person, location and organization respectively on the People's Daily on January 1998.