摘要
提出了一种基于实例和错误驱动相结合的规则学习方法。该方法首先将提取的文本中的语法结构信息作为实例,然后采用基于转换的错误驱动学习方法找出这些实例的适用上下文环境,从而建立相应的规则库。此方法提取出的规则完全采用机器学习的方式,避免了人工提取规则的主观性缺点。可用于诸如词性标注、未登录词识别、命名实体抽取等自然语言研究课题。
A new rule learning method based on the instance and error-driving is proposed. Grammar structure information is extracted and taken as the instance,and then the context of the instance is found according to the error-driving learning method. Thus,a rule-base is constructed. The model of machine learning is adopted ,and the subjectivity default caused by manual work is avoided. The method could be used in many natural language research subjects, such as unknown word recognition ,part of speech tagging and name entity extraction ,etc.
出处
《计算机应用与软件》
CSCD
北大核心
2008年第1期162-164,共3页
Computer Applications and Software
关键词
规则学习
中文信息处理
专有名词识别
Rule learning Chinese information processing Proper noun recognition
