Artificial Intelligence-based Recognition of Chinese Characters and Translation of Classical Chinese: Focusing on the Current State of Korea

Abstract

This paper introduces the current status of artificial intelligence-based recognition of Chinese characters and machine translation of Chinese characters (Classical Chinese texts), which is currently underway in Korea, and suggests future tasks. For the recognition of Chinese characters, this paper discusses the Chinese character segmentation using the Mask R-CNN algorithm and the Chinese character classification method using the Inception- Resnet-V2 algorithm, and explores the significance and limitations of the construction of Chinese character forms database. Regarding machine translation, this paper looks at the development of machine translation technology from Rule-Based Machine Translation (RBMT) to Neural Machine Translation (NMT); then, introduces the principles of neural network machine translation and techniques applied to improve performance. There are two types of machine translation for Classical Chinese texts: the machine translator provided by Baidu and the translator developed by the Institute for the Translation of Korean Classics (ITKC). Although the Baidu machine translator has various errors, the overall translation performance is somewhat higher than that of ITKC's machine translator. The ITKC's translator performs better than the Baidu translator in certain areas, but falls short of the Baidu translator in translating various types of Classical Chinese texts. In order to improve the performance of the machine translation of Classical Chinese texts, it is necessary tobuild a cloud-based translation platform, expand a high-quality parallel corpus, develop an additional system to improve machine translation performance, and develop a highly reliable translation evaluation method.