Abstract: Syntactic analysis is one of important constituent parts in the field of natural language processing. An neural network-based Generalized LR(NNGLR) syntactic analysis algorithm was described in this paper. The algorithm unites GLR algorithm with neural network, and the shift-reduce parsing decision of GLR parser was simulated by a back-propagation neural network so as to improve its flexibility. Based on the techniques above, the algorithm solved the disadvantage of GLR that parsing table becomes large when there exists many shift-reduce conflicts. The experiment shows that the algorithm has good generalization.

Key words: neural network, GLR algorithm, syntactic analysis, context-free grammar

摘要： 提出并实现了一种基于神经网络的GLR(GeneralizedLR)句法分析算法,该算法结合神经网络自学习、自组织和并行分布处理等优点,以BP神经网络结构模型取代了GLR算法的分析表,模拟其移进和归约动作,通过计算网络输出来分析句法结构。该分析算法较好地解决了GLR算法对于存在多个移进归约冲突动作时,复制分析栈会使得动作表变得很大的缺点,实验结果表明,这种算法具有较好的泛化能力。

关键词: 神经网络, GLR算法, 句法分析, 上下文无关文法