한국어·외국어 병렬 말뭉치 구축의 쟁점

An Issue of Korean-Foreign Language Parallel Corpus Construction: Focusing on Original Text Refinement

This study introduces the project of the Korean-Foreign language parallel corpus conducted by the National Institute of the Korean Language and examines the contents related to original text refinement as a process for advanced quality in corpus construction. The process of original text refinement is supremely important in the process of constructing a parallel corpus on the premise of foreign language translation. Since the translator who must perform the translation is not a Korean but a native speaker of the target language, errors or mistakes in the original text are too difficult to guess and understand the context. Therefore, establishing the principles for original text refinement and having consistency can be a great help in enhancing the translation quality of the corpus. The target languages for parallel corpus construction are eight languages (Vietnamese, Indonesian, Thai, Hindi, Khmer, Tagalog, Russian, and Uzbek). In this research, three stages of refinement such as mechanical refinement, refinement at a non- expert group level and expert group level refinement were conducted to advance quality in translation. And the original texts were modified in terms of accuracy and normativity, in terms of publicness and ethics and in terms of clarity of information and ease of translation while the minimum scope of modification was made to avoid the damage of the meaning. In spoken Korean, however, it needs to be considered regarding whether it would be a more desirable direction to literally translate the language phenomena used by the actual public. Especially, they conform to norms but there is a need for more extensive discussion on how to handle the loanwords which are unused by the actual public.