结合全局向量特征的神经网络依存句法分析模型  

Neural network model for dependency parsing incorporating global vector feature

作　　者：王衡军[1] 司念文 宋玉龙 单义栋 

机构地区：[1]解放军信息工程大学三院 [2]73671部队

出　　处：《通信学报》2018年第2期53-64,共12页Journal on Communications

摘　　要：利用时序型长短时记忆(LSTM,long short term memory)网络和分片池化的卷积神经网络(CNN,convolutional neural network),分别提取词向量特征和全局向量特征,将2类特征结合输入前馈网络中进行训练;模型训练中,采用基于概率的训练方法。与改进前的模型相比,该模型能够更多地关注句子的全局特征;相较于最大化间隔训练算法,所提训练方法更充分地利用所有可能的依存句法树进行参数更新。为了验证该模型的性能,在宾州中文树库(CTB5,Chinese Penn Treebank 5)上进行实验,结果表明,与已有的仅使用LSTM或CNN的句法分析模型相比,该模型在保证一定效率的同时,能够有效提升依存分析准确率。LSTM and piecewise CNN were utilized to extract word vector features and global vector features, respec-tively. Then the two features were input to feed forward network for training. In model training, the probabilistic training method was adopted. Compared with the original dependency paring model, the proposed model focused more on global features, and used all potential dependency trees to update model parameters. Experiments on Chinese Penn Treebank 5 （CTB5） dataset show that, compared with the parsing model using LSTM or CNN only, the proposed model not only re-mains the relatively low model complexity, but also achieves higher accuracies.

关 键 词：依存句法分析 图模型 长短时记忆网络 卷积神经网络 特征 

分 类 号：TN912.3[电子电信—通信与信息系统]