The Intelligent Archives Service Platform is an intelligent service platform for screening classified documents and reviewing archives for opening to the public. To deal with archival documents that mix handwriting, printed text, tables and seals, a dedicated OCR image recognition model was developed in-house. It accurately recognizes all types of elements and restores the original text of the archives. By studying a large number of cases of classification markings, the team built a classification marking dataset and optimized the model for it, which greatly improved the ability of the model to recognize the markings of classified documents. In addition, to accurately identify classified archives that have no markings, the archive content must be checked and understood in full from a semantic perspective. Based on the natural language understanding ability of pre-trained models, the platform parses and compares archival texts to identify classified files and sensitive content.