A Quality Prediction Model for the Parallel Corpora of Korean-Chinese and Chinese-Korean Translation Utilizing Sentence Similarity and Sentence Attributes

This study proposes a model for assessing the quality of Korean-Chinese and Chinese-Korean parallel translation corpora and applies it to assess the quality of existing Korean-Chinese and Chinese-Korean parallel translation corpora. The significance of the study is, firstly, it conducted Quality Estimation (QE) of Korean-Chinese and Chinese-Korean translation corpora using Direct Assessment (DA) scores, establishing related data for the first time. Secondly, it established a model for assessing the quality of parallel translation corpus, called TwiQE. Thirdly, it provided the basis for improving the quality of the parallel translation corpus by presenting quality standards established by multiple experts. Finally, using this model, it conducted a quality assessment on the two types of Korean-Chinese and Chinese-Korean parallel translation corpora available on AI HUB. Based on the quality assessment scores by TwiQE, the corpora that needed improvement was classified into two levels, and the quantity and proportion of corpora that did not meet the standards for each level to provide a basis for improving the quality of the corpora identified.