Students from Data Science Department Publish Scientific Paper at RIVF 2023 International Conference

Wed, 25/10/2023 - 23:23

Paper Title: "OCR-Based Information Extraction on Vietnamese Book Cover Images"

Authors:

Nguyen Minh Tam - 20520748 - DS2020 - Primary Author

Nguyen Truong Thinh - 20520783 - DS2020 - Co-Author

Tran Pham Gia Phu - 2050694 - DS2020 - Co-Author

Quach Co Thai - 20520756 - DS2020 - Co-Author

Supervisor: Ph.D. Do Trong Hop

Abstract:

In modern libraries, the systematic digitization of books for storage and structured information retrieval is becoming increasingly important, especially as the number of published books in the market continues to rise rapidly. Traditionally, information related to books, such as titles, authors, publishers, etc., would be manually typed by staff in libraries or bookstores, leading to significant waste of time and labor. With the rapid development of machine learning and deep learning technologies, computers can now assist us in performing tasks of recognition and classification of information quickly, accurately, and efficiently. Specifically, Optical Character Recognition (OCR) technology and Natural Language Processing (NLP) can automate the process of digitizing books, significantly saving labor costs and improving efficiency. Text on book cover images is recognized by an OCR model before being extracted into specific fields such as titles, authors, publishers, etc., using classification models.

In this paper, the team focused on extracting book information from text on book cover images in Vietnam using the Vi-BCI dataset comprising over 7000 labeled book cover images. The team built various models for the task from different approaches, all of which showed high predictive results. The team selected the best-performing models from each approach and evaluated their results using the Character Error Rate (CER) metric. From these results, the team analyzed mispredictions to find remedies and proposed future development directions.

We extend our heartfelt thanks to Dr. Do Trong Hop for his guidance and support throughout our research journey and the publication of this international scientific paper. We also express our gratitude to Ms. Nguyen Thi Tham and Ms. Vo Kieu Hoa, former students from the DS2018 class of our university, for allowing us to use the Vi-BCI dataset and providing valuable insights during our research process.

The RIVF conference is an international event in the fields of Information and Communication Technology and Computing. It serves as a major scientific event bringing together researchers and scholars in computer science and communication technology in Vietnam and the world, following the theme "Research, Innovation, and Vision for the Future" (RIVF). The RIVF conference is listed among the prestigious conferences as proposed by SCOPUS and ISI Web of Science. RIVF has gone through 16 editions, and 2023 marks its 17th edition. It is one of the highly reputable scientific conferences in the field of Information and Communication Technology and Computer Science. The conference is scheduled to be held in Hanoi in December 2023.