Literary machine translation: The case of Danish cultural references

This study examines how Google Translate, a neural machine translation system, handles cultural references when translating excerpts from a Danish novel into Spanish and English. Using a qualitative analytical framework combining cultural classification categories with translation structure categories and cultural reference equivalence, we analysed 24 segments across five categories: physical/material objects, gestures/habits, social roles/personalities, social organisations/businesses, and religious/social customs. Our findings reveal that Google Translate produces predominantly literal translations in both language pairs, but linguistic proximity significantly affects outcomes. Google translations into English achieved considerably more cultural reference preservation compared to Spanish, while Spanish translations showed higher rates of cancellation and foreignising. These results highlight persistent challenges when culturally embedded references are machine translated, particularly when cultural meaning is conveyed implicitly. The study has implications for translator training and the Danish publishing industry, and underscores the need for reader reception research in literary machine translation.