关键词: 智能问诊, 反问生成, 文本循环神经网络, 双向变形编码器, 开源神经机器翻译

Abstract: Abstract: In order to solve the problems such as time consuming, inaccuracy and influence of parameter setting on algorithm performance when optimizing parameters of convolutional neural network by traditional manual methods, a variable convolutional autoencoder method based on teaching-learning optimization was proposed. The algorithm designed a variable length individual encoding strategy, quickly constructed the convolutional autoencoder structure, and stacked the convolutional neural network. In addition, the excellent individual structure information was fully utilized to guide the algorithm to search more promising regions and improved the algorithm performance. The experimental results show that the proposed algorithm achieved a classification accuracy of 89.84% when solving medical image classification problems, which was higher than that of traditional convolutional neural network and similar neural network. The algorithm solved the problem of medical image classification by optimizing the structure of the convolutional autoencoder and stacking the convolutional neural network, which effectively improved the performance of medical image classification.

Key words: intelligent consultation, rhetorical question generation, Text Recurrent Neural Networks(TextRNN), Bidirectional Encoder Representations from Transformers(BERT), Open-Source Neural Machine Translation(OpenNMT)