基于转移的神经网络哈萨克语句法分析

为了进一步提高哈萨克语句法分析的准确率，为哈萨克语自然语言处理奠定良好基础，对基于转移的哈萨克语句法分析进行研究，采用改进后的基于转移的方法对句法树进行处理，即中序遍历句法树的方法将句法树转换为动作序列。使用神经网络构建句法分析器框架，分别使用三个长短期记忆网络（LSTM）表示堆栈信息、缓冲区信息以及动作历史信息对模型进行训练，根据所得到的概率预测动作序列，从而得到句法分析的结果。改进后的转移方法得到的句法分析准确率为74.37%。

Transition-Based Kazakh Parsing with Neural Network

For purpose of improving the parsing accuracy of Kazakh and laying the foundation for natural language processing, it researches Kazakh parsing based on transfer, and uses an improved transition-based method to deal with the syntax tree and convert the syntax tree into an action sequence, this method is in-order traversal over syntactic trees. The neural network is used to construct the parser framework, and three long short-term memory are used to express the stack information, buffer information and action history information to train the model. According to the probability of predicting the sequence of action, the result of syntactic analysis is obtained. The accuracy of Kazakh parsing obtained by the improved transition-based method is 74.37%.