基于循环神经网络的工程专业语义智能分析方法研究

Research on engineering semantic intelligence analysis method based onRecurrent Neural Network

针对传统翻译方法学习能力差、翻译质量较低的问题,提出了一种基于循环神经网络的专业英语机器翻译方法。该方法以编码器-解码器为模型框架,利用改进的循环卷积神经网络对输入数据加以训练。同时编码器使用多头注意力机制对输入数据进行共同训练,进而使算法兼具局部与全局特性。解码器单层则采用三子层结构,分别为多头注意力子层、上下文信息子层及全连接子层,可保证句子翻译的流畅性。在实验测试中,所提算法的BLEU值与其他算法相比提升了2.7;而在专业语料翻译测试中,相较于网络翻译,该算法的准确性和流畅度均更优,由此表明其性能较好,具有一定的工程应用价值。

To solve the problem of poor learning ability and low translation quality of traditional translation methods,a method of professional English machine translation based on Recurrent Neural Network is presented.The algorithm uses the encoder⁃decoder as the model framework,uses the improved recurrent convolution neural network to train the input data,and uses the multi⁃head attention mechanism to train the input data together,so that the algorithm has both local and global characteristics.The decoder uses three sublayers,namely,multi⁃head attention sublayer,context information sublayer and fully connected sublayer,to ensure the smoothness of sentence translation.In the experimental test,the BLEU value of this algorithm is improved by nearly 2.7 compared with other algorithms.At the same time,in the professional corpus translation test,the accuracy and fluency of this algorithm are better than those of network translation,which indicates that the performance of this algorithm is better and has certain application value.