摘要
中文机构名自动识别是自然语言处理的一个难点,提出一种基于条件随机场的字词模型相结合的识别组织机构名的方法。该方法针对机构名的特点并利用知网进行两方面的特征选择。在开放测试中,字模型和词模型的F-值分别为91.51%和91.09%,两者进行互补结合之后F-值分别为92.37%和92.06%,说明字词模型结果具有互补差异性,相结合可以取得比单一模型更好的结果。
Automatic recognition of Chinese organization name is a very difficult problem in natural language processing. This paper presents a new method for Chinese organization name recognition based on the cooperation of character model and word model using conditional random fields. This method selects two kinds of fea- tures respectively from the feature of the organization name and Hownet. In open test, the F - value of the character model and the word model are 91.51% and 91.09% , it obtains 92.37% and 92.06% after com- bined them. It suggested that the combined model is performing better than each single model.
出处
《沈阳航空工业学院学报》
2009年第1期49-52,共4页
Journal of Shenyang Institute of Aeronautical Engineering
关键词
机构名自动识别
字模型
条件随机场
知网
automatic recognition of organization name
character model
conditional random fields
hownet
