基于深度学习的英语口语自动评分研究

Research on automatic oral English scoring based on deep learning

为提高英语口语学习效率,在研究了深度神经网络、声学模型、自动语音评分基础上,设计了一种基于深度学习的端到端自动英语评分系统。在输入层,将词汇表示为一个序列张量,其中每个位置对应于来自预先训练的词汇向量,并采用双向LSTM网络得到高层信息。在声学模型层,文中将注意力机制模型融入网络,从而提高系统运行效率。在输出层,将词汇的表示方式与口语表达连接起来,并利用softmax函数对成绩进行预测。通过仿真分析,结果表明所提方法较传统LSTM、GRU方法性能有所提升。仿真结果进一步验证了所提系统的有效性。

In order to improve the efficiency of oral English learning,an end-to-end automatic English scoring system based on deep learning is designed on the basis of deep neural network,acoustic model and automatic voice scoring.In the input layer,the vocabulary is represented as a sequence tensor,in which each position corresponds to the vocabulary vector from the pre-training,and the Bi-directional LSTM network is used to obtain the high-level information.In the acoustic model layer,the attention mechanism model is integrated into the network to improve the efficiency of the system.In the output layer,the expression of vocabulary is connected with the oral expression,and the softmax function is used to predict the performance.Simulation results show that the performance of the proposed method is better than that of the traditional LSTM and GRU methods.Simulation results further verify the effectiveness of the proposed system.