A Novel Extraction Method with Document-Level Relationship Based on Big Data and Pre-Trained Models

In recent years, due to the increasing demand for document-level relationship extraction in industrial enterprises, is necessary to build a document-level relationship extraction platform based on big data and pre-trained models. The platform can extract information based on the analysis of operators' information extraction behavior, and realize the relationship extraction function, thereby improving the operation effect of staff. In the research process, based on big data and pre-trained models, this paper promotes the planning of each module of the relationship extraction platform, extracts the corresponding coordinate information of document relationships with the help of OCR tools, improves and optimizes the functions of complex layout, pre-training, input indicators, document sorting, information extraction, relationship recognition, etc., and then carries out the fusion processing of document-level relationship extraction, and puts it into the application of the platform. The results of this paper show that in practical application, the recognition accuracy of the platform is 88.36%, the extraction accuracy of the operator is increased by 20.54%, and the average improvement of the layout level is 23.99%. It can be seen that document-level relationship extraction based on big data and pre-trained models can significantly improve the operational efficiency and quality of industrial enterprises.