Method for solving related full-text interpretation and analysis problems

The invention discloses a method for solving related full-text interpretation and analysis problems. The method comprises the following steps of 1, data preprocessing; 2, model construction; and 3, result integration. According to a solution idea of the invention, a neural network with multiple input ends is introduced, different problems and document contents are fused by using an appropriate algorithm, various defects caused by single use of training are avoided by using a method of combining rules with training, and finally, answers obtained by two different means of an extraction method and a generation method are integrated by using a reasonable integration algorithm. 本发明公开了一种解决有关全文解读分析问题的方法，包括以下步骤：步骤一：数据预处理：步骤二：构建模型：步骤三，集成结果。本发明一种解决思路，通过引入多输入端的神经网络，使用合适的算法融合不同问题和文档内容，并使用规则结合训练的方法，规避单独使用训练的各种弊端，最后使用合理的集成算法将抽取式方法和生成式方法这两种不同手段获取的答案集成。