摘要
论文提出了一种基于BP网络的汉语句法分析错误恢复方法,结合神经网络自学习、自组织的优点,以神经网络的结构模型代替了富田胜算法的分析表,模拟其移进规约动作。该方法鲁棒性好,能够恢复富田胜算法句法分析结果中的若干错误句子,提高其容错性能。实验结果表明,该方法对于错误句子的恢复取得了令人满意的效果。
An error recovery method of Chinese syntactic analysis based on neural network is described in this paper. GLR algorithm is united with merits of neural network,in which the shift-reduce parsing decision of GLR parser is simulated by a back-propagation neural network so as to improve its flexibility,robust parsing is thus achieved.The method can analyze umpty error sentence from results of GLR syntactic analysis.The experiment shows that the proposed algorithm obtains a satisfactory effect in recovery of error sentence.
出处
《计算机工程与应用》
CSCD
北大核心
2005年第31期74-75,78,共3页
Computer Engineering and Applications
关键词
神经网络
富田胜算法
汉语句法分析
错误恢复
neural network,GLR algorithm,Chinese syntactic analysis,error recovery
作者简介
赵亚琴(1973-),女,山西原平人,博士研究生,主要研究方向为信息系统工程。智能信息处理。周献中(1962-),男,教授,博士生导师,主要研究方向为智能信息处理,信息系统工程,指挥自动化。