Intelligent identification and classification of requirement document content based on CNN and RNN

With the speedy prosperity of artificial intelligence, computer-based deep learning methods have been commonly used in document recognition and classification. Because the traditional convolutional neural network ignores the details of the document when identifying the document content, it cannot obtain the specific information of the document image content. Therefore, a two-branch convolutional neural network model combining convolutional neural network and recurrent neural network is developed to identify and classify the required document content accurately and efficiently. The outcomes indicate that the accuracy of the raised model is 12.62% and 11.09% higher than that of the bidirectional long short-term memory network and convolutional recurrent neural network, respectively. When the number of iterations is 800, the raised model converges and its loss function is 0.001. Under the SST and SSGT data sets, the accuracy of the raised model is 0.996 and 0.994, respectively, which is 1.55% and 0.93% higher than that of the spatial feature extraction branch and multi-layer feature fusion branch models, separately. The outcomes indicate that the raised model has better performance for document content recognition and classification. This research is important for enhancing the retrieval efficiency and recognition accuracy of document content.