A Study on Process Method of Modal Adverbs in AI Translation : Focusing on Japanese-Korean Translation of Novels

The processing patterns of modal adverbs that appear in AI translation were examined by dividing them into Korean correspondence patterns and translation processing patterns. The results are as follows. First, in terms of Korean response, ‘もちろん/やはり/なぜ/まったく/まるで｣’(Google and Papago) and ‘きっと/おそらく/たぶん’(Papago) show a one-on-one response to Korean at a rate of more than 90%. Particularly, ‘もちろん/まるで/まったく/やは り’(Google) and ‘なぜ/おそらく/たぶん/やはり’(Papago) are translated only to [물론/마치/전혀/역시] and [왜/아마(도)/아마(도)/역시], which means the overall semantic function is not subdivided and the diversity of expression is insufficient. Second, the world order change was the least, and the omission was the most common in terms of translation processing. The fixed adverbial position shows the weakening of the floating characteristic of the adverb, and the omissions show the adverbial feature as a secondary component. Subsequently, the misuse due to the confusion of the positive and negative usage methods of ‘まるで’, transformation due to the parts of speech transformation of ‘な ぜ’(Google) and ‘もちろん’(Papago), and meaning transformation of ‘きっと’ (Papago) can be seen. Misuse appears 36 on Google(1.86%) and 4 on Papago (0.20%), and transformation appears 1 on Google(0.05%) and 8 on Papago(0.41%). It shows flexibility partially with accuracy of machine translation.