A comparative analysis of human and ChatGPT translations in an English-Korean parallel corpus: Focusing on animacy encoding in referential expressions

This study examines how referential expressions in the English source text of Shakespeare’s Hamlet are realized in Korean human and ChatGPT translations regarding animacy encoding between English and Korean. Focusing on the main protagonist Hamlet, his father, and his uncle, it compares the frequency and syntactic distribution of pronouns, proper noun phrases, and common noun phrases across the three texts by chi-square analysis and qualitative examination. The English text favors pronouns while both Korean translations prefer proper noun phrases. In particular, the ChatGPT version shows greater syntactic variation, omissions, and reduced lexical diversity, whereas the human translation more sensitively adapts referential choices to the discourse context, underscoring both the potential and the limitations of ChatGPT for discourse-level translation of referential expressions. The study extends prior sentence-level analysis by assessing ChatGPT translation quality at the discourse level.