The purpose of this thesis is both to take a review of the errors in English-Korean translation by AI translators(machine translation) and to examine the characteristics of Al translators related to the errors types.
The source texts used in this thesis are non-literary texts such as editorials and columns. We compared and analyzed the outcomes of two AI translators, Google Translate and Naver Papago.
We studied four grammar categories: pronouns, plural markers, referents and postponement. In chapters two and three, we focused on how Google translation and Papago handle pronouns and plural markers. In chapters four and five, we focused on whether Google and Papago correctly identify the referents of pronouns and postponement.
The results are as follows: according to the results of the analysis of chapters two and three, there are distinct differences in the translations produced by Google and Papago. Google translate usually has shown a tendency to follow the rules of Korean grammar, while Papago has shown a tendency to follow those of English grammar. According to the results of chapters four and five, it is hard to find a distinct difference between AI translators. However, we can find a difference between human and AI translators. Unlike English and Korean speakers, AI translators have a problems in correctly grasping and applying grammar rules.
From the above results, this thesis has the following implications; 1) Google Translate tends to focus on Korean grammar rules and Papago tends to translate with an emphasis on English grammar. Based on these results, it can be inferred that Source-Oriented translation strategy is a feature of Google translate, while Target-Oriented translation is a feature of Papago. It can be also inferred that the native language environment of the developer may have influenced the translation tendency of the AI translators. 2) AI translators have difficulty in translating specific grammar categories.
Among this thesis's contributions is an examination of the performance of AI Translators that was able to identify their respective unique characteristics. It also has practical implications. We expect that the research results of this thesis will be useful information for users of AI translators as well as for human translators. In addition, the research method of this thesis, which is analyzing the AI translator's errors and grammatical differences between languages, can be utilized in the field of language education.