Development of an AI System for Data Extraction from Vietnamese Printed Documents
Phuc Huynh Vinh, Phi Nguyen Xuan, Ouan Chu Nhat Minh, Duc Bui Ha · 2024
In Vietnam, most official documents such as legal contracts, licenses, and application forms still exist in traditional paper-based formats. These paper-based documents are not accessible and expensive to maintain. Hence, there is a significant demand for the digitization of information from these documents. This study aims to develop a system that can analyze and filter information from paper documents accurately and efficiently by employing artificial intelligence (AI) techniques. The system is built upon a combination of deep learning models, including YOLO for document layout analysis and document parsing and TransformOCR for text recognition, which is refined to handle handwriting text. Various document pre-processing pipelines were investigated to enhance the accuracy and efficiency of information extraction. The experimental results demonstrate that the YOLO model achieved a precision of 91% in layout analysis and 89% in parsing, while the TransformerOCR model reached CER of 12.24% and WER of 36.04% in text recognition task.