Page 123 - Kỷ yếu hội thảo khoa học quốc tế - Ứng dụng công nghệ mới trong công trình xanh , lần thứ 8

106 TRƯỜNG ĐẠI HỌC SƯ PHẠM KỸ THUẬT - ĐẠI HỌC ĐÀ NẴNG

Application of Optical Character

Recognition Method: A Case Study for

Educational Information System

Cong-Doan Truong Tran Viet Long Ngo Quang Truong
*
International School International School International School
Vietnam National University Vietnam National University Vietnam National University
Hanoi, Vietnam Hanoi, Vietnam Hanoi, Vietnam
tcdoan@vnu.edu.vn tranvietlong2k@gmail.com 19071638@vnu.edu.vn
Ha Manh Hung Hoang Huong Chi Ninh Thi Linh
International School International School International School
Vietnam National University Vietnam National University Vietnam National University
Hanoi, Vietnam Hanoi, Vietnam Hanoi, Vietnam
hunghm@vnuis.edu.vn 22070449@vnu.edu.vn 19071612@vnu.edu.vn
Pham Duy Duong Satyam Mishra Do Cong Tuan
Faculty of Electrical and Electronic International School International School
Engineering Vietnam National University Vietnam National University
The University of Danang - University Hanoi, Vietnam Hanoi, Vietnam
of Technology and Education satyam.entrprnr@gmail.com 19071639@vnu.edu.vn
Danang, Vietnam
pdduong@ute.udn.vn

Abstract: This research presents a multi-lingual Optical reliable OCR performance independent of network
Character Recognition (OCR) system designed to extract connectivity. User interfaces were designed to be language-
text from international student ID cards, catering to English agnostic. Future work includes expanding language
and Vietnamese languages. The multi-lingual OCR model coverage, increasing training data, and exploring advanced
was compared to individual uni-lingual models, neural network architectures. Integration with student
demonstrating comparable accuracy, with a modest 1-3% information databases and administration systems for
decrease. Key error analysis revealed challenges in streamlined access and real-time language translation for
character segmentation and language disambiguation, enhanced usability are also suggested. In summary, this
particularly in cursive scripts. To enhance accuracy, research lays the foundation for inclusive OCR systems in
optimization measures including increased training data, international universities, offering equitable experiences for
regularization, CTC loss training, ensemble modelling, and diverse student populations and facilitating administrative
on-device dictionaries were proposed. The study achieved processes.
an average accuracy of 94.5% for the multi-lingual model Keywords: Optical Character Recognition (OCR), Multi-
and 95.35% for the uni-lingual model, showcasing the lingual Text Recognition, Deep Learning, Mobile
feasibility of this approach despite the complexities of Deployment, Language Translation, Machine Learning
diverse scripts and designs. The system was efficiently
optimized for on-device deployment, ensuring fast and
I. INTRODUCTION more efficient process in generating and managing ID
cards, leading to a more robust solution and a more
Existing systems for student ID cards in Vietnam efficient delivery of service [4]. Additionally, the
have limitations in terms of content, format, and test invention of an intelligent electronic student ID card
administration, which negatively impact the with a control module, charging module, positioning
communication goal of foreign language education in module, and communication module offers automatic
Vietnamese schools [1]. These limitations call for a and accurate attendance recording, real-time student
radical renovation in both the format and position detection, and improved positioning
administration of the national matriculation and accuracy [5].
general certificate of secondary education English test
[2]. On the other hand, electronic ID systems in The objective of this research is to develop and
educational institutions have advanced technologically, evaluate a multi-lingual optical character recognition
but there is a need for enhancements to improve their (OCR) system capable of extracting text from
capabilities [3]. These enhancements aim to achieve a international student identification cards. While
existing systems such as the one described in our
ISBN: 978-604-80-9122-4