摘要
利用条件随机场能够任意添加特征的优点,基于条件随机场模型在字粒度一级进行中文人名识别的研究.根据中文人名在文本中出现的基本特征和上下文特征,结合模型的综合性能,合理构造条件随机场的特征模板,在大规模标注语料上进行训练,统计中文人名在文本中的条件概率分布,获得模型参数,并采用序列标注的方式完成中文人名识别任务.多次闭合测试和开放测试结果表明,F值基本都高于90%.
Taking advantage of the ability of using arbitrary features as input in CRFs, the task of the name of Chinese people recognition was discussed based on the Conditional Random Fields on the character level. According to the basic and context features of the Chinese people＇s names, the feature template of CRFs was built reasonably combining the comprehensive performance of the model which was trained on the large scale corpus to obtain the model＇s parameters by counting the Chinese names＇ conditional probability distribution in the texts. By sequence labeling, it implemented the recognition of Chinese names. It obtained promising results on different closed and opened test corpus with the F measurement value of almost 90% above.
出处
《昆明学院学报》
2011年第6期64-66,共3页
Journal of Kunming University
基金
昆明学院科研课题资助项目(2009G012)
关键词
命名实体识别
中文人名识别
条件随机场
条件概率
特征模板
序列标注
named entity recognition
Chinese people＇ s name recognition
conditional random fields
conditional probability
feature template
sequence labeling
