摘要
情感分析方法从浅层学习不断地向深度学习探索。在深度学习当中,循环神经网络(RNN)由于学习层数的加深,会导致一定的梯度弥散的问题,由此人们提出一种长短时记忆(LSTM)网络模型解决该问题。提出一种神经网络架构模型——Multi GRU,采用LSTM的一个重要变种GRU(Gated Recurrent Unit)进行模型构建。Multi GRU利用GRU进行多层堆叠,以减少信息丢失。实验论证该模型在损失等方面的表现比LSTM等模型更好。
Sentiment analysis methods continuously explore from shallow learning to deep learning.In the deep learning, recurrent neural network（RNN） due to the deepening of the number of learning layers, leads to a problem of vanishing gradient. Thus, long short-term memory（LSTM） network model was proposed to solve the problem. We proposed a neural network architecture model： MultiGRU. The model used LSTM’s an important variant GRU. MultiGRU used GRU for multi-layer stacking to reduce information loss. Experimental results show that the model is better than LSTM in terms of loss and so on.
作者
邢长征
李珊
Xing Changzheng,Li Shan(1.School of Electronics and Information Engineering, Liaoning Technical University, Huludao 125105, Liaoning, Chin)
出处
《计算机应用与软件》
北大核心
2018年第8期102-106,共5页
Computer Applications and Software
