Numerical data acquisition system, neural network full-connection model and training method

The invention provides a numerical data acquisition system, a neural network full-connection model and a training method, and the method is characterized in that the method comprises the steps: extracting numerical information in training data; labeling the numerical value information in a preset category to obtain a labeling sequence; inputting the labeling sequence into a feature sequence matrixoutput model to obtain a feature sequence matrix; and based on the feature sequence matrix, performing reverse iterative training for multiple times by a neural network full-connection model. Throughthe training method provided by the invention, the neural network full-connection model with the numerical semantic analysis capability can be obtained, specific numerical information in the legal text can be accurately obtained by utilizing the neural network full-connection model, and the numerical extraction capability in a new text in the legal field is improved. 本发明提供了数值数据的获取系统、神经网络全连接模型及训练方法，其特征在于，包括：提取训练数据中的数值信息；对所