News content label automatic extraction method and system

• Main Author: ZHANG MAOHONG
• Format: Patent
• Language: Chinese; English
• Published: 14.05.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention relates to a news content label automatic extraction method and system, and the method comprises the steps: a sample data obtaining step for obtaining sample data; a sample label labeling step for constructing a label system and performing label labeling on the sample data based on the label system to obtain a training sample; a model construction step for forming a fusion neural network model based on a BERT network, a TextCNN network and a capsule network, and training the fusion neural network model as a label automatic extraction model by using the training sample; a model prediction step for inputting the news content to be processed into the label automatic extraction model to obtain the label of the news content to be processed. The development cost and the time cost are reduced, and the model prediction accuracy is improved. 本申请涉及一种新闻内容标签自动抽取方法及系统，其中，该新闻内容标签自动抽取方法包括：样本数据获取步骤，用于获取样本数据；样本标签标注步骤，用于构建标签体系并基于所述标签体系对所述样本数据进行标签标注，得到训练样本；模型构建步骤，用于基于BERT网络、TextCNN网络及胶囊网络形成一融合神经网络模型，并利用所述训练样本训练所述融