APPLICATION OF BERT ARCHITECTURE FOR STORAGE TIME OF RECORD CLASSIFICATION PROBLEM

Abstract

Record storage at the competent agencies and organizations is an essential problem in the management and organization of document preservation. However, with the increasing number of archives and many different types of documents, leading to overloading documents during the archiving process. Therefore, the classification of records according to the preservation period is a very important step in preservation, contributing to optimize the composition of the archive fonts, and save the cost of document Therefore, in this paper, we present a study evaluating the effectiveness of the BERT model compared with traditional machine learning and deep learning algorithms on a real-world dataset to solve this task automatically. Experimental results show that the BERT model achieved the best results with 93.10% of precision, 90.68% of recall and 91.49% of F1-score. This result shows that the BERT model can be applied to build systems to support record classification in the real-world application is completely feasible.