일본의 한적 정보화와 코퍼스 구축 현황

Abstract

In the case of Korean domestic resources for classic literature written in Chinese characters, the government-driven project for digitization of classical texts has secured the raw corpus, but the development of tagged corpus and parallel corpus has not received as much attention.The development of corpus and the machine translation of Korean domestic Chinese character text resources were not a phased development over time, but rather a short-term advancement helped by recent developments in artificial intelligence translation. Artificial intelligence translation at the current stage is processed by machines automatically learning the relationship between sentences based on the data of original and translated texts, and therefore developing a tagged corpus, which is adding annotations according to certain criteria such as grammar, word class, and punctuation, is deemed an unnecessary task. In comparison, the situation is very different in China and Japan, which are also within the Chinese character cultural sphere. China and Japan have been phasing in with projects in digitalization of texts in Chinese characters and development of corpora in preparation for machine translations. The goals of developing corpora in China and Japan is not only to improve the efficiency of machine translations, but also to positively contribute to the national studies of each country. This situation prompts us to reconsider whether it is truly unnecessary to develop corpora for Korean texts in Chinese characters. In Japan, Kyoto University Institute for Research in Humanities Library, and the National Institute for Japanese Language and Linguistics have led the corpus development work from original texts in Chinese characters. The Institute for Research in Humanities have developed a tagged corpus utilizing the Mecab Morphological Analyzer and attempted dependency parsing based on Universal Dependencies with application of the traditional Japanese reading method(Kundoku) of Chinese characters. The studies of the National Institute for Japanese Language and Linguistics not only created translation results by developing corpora, but also extracted scholarly, literary styles depending on the era, the schools of theory, the literary circles in Japan from this foundation. Considering such examples, the development of tagged, parallel corpora from Korean original text resources in Chinese characters is still a meaningful task.