발간년도 : [2022]
| 논문정보 |
|
| 논문명(한글) |
[Vol.17, No.6] PaperCleaner, an Automatic Restoration Program of Korean Documents Based on GAN Algorithm |
|
| 논문투고자 |
Min-Kyu Kim, Hyoung-Taek Lim, Gyesik Lee |
|
| 논문내용 |
This paper describes a tool, called PaperCleaner, for the automatic restoration of Korean documents. The usefulness of PaperCleaner can be recognized when there are lots of digital papers in PDF or Jpg format whose original versions should be recovered. The need for such work would be especially critical for people working in educational institutions. Teachers for example would like to erase all the personal notes written on test papers before they can be distributed to students for exercises. But currently no applications or programs are known which automatically remove personal notes and recover the original test papers. PaperCleaner does the work without any human intervention. The main content of this paper is to introduce the main idea of our tool and to demonstrate its performance. We used our tool to restore the original test papers containing questions, answers, and some personal notes. The main algorithm is based on a Generative Adversarial Network algorithm (GAN). GAN is a class of machine learning frameworks introduced in 2014. The idea of the algorithm is that two neural networks contest with each other in the form of a zero-sum game. One agent’s gain means another agent’s loss. PaperCleaner is very simple to use. If a document is submitted as an input to our tool, then a restored document will be produced where almost all personal notes are removed. |
|
| 첨부논문 |
|
|
|
|
|
자료실 > 논문지 
