Fluency and Faithfulness in Human and Machine Literary Translation

TL;DR

This study investigates the relationship between fluency and faithfulness in literary translation, revealing a tradeoff where increased fluency often correlates with decreased semantic faithfulness, especially in machine translation.

Contribution

It introduces a large-scale analysis of literary translation, highlighting the tradeoff between fluency and faithfulness across human and machine translations.

Findings

Negative correlation between fluency and faithfulness in translations.

Pattern consistent across human and Google Translate, weaker in TranslateGemma.

Segment length influences automatic evaluation metrics.

Abstract

Literary translation requires balancing target-language fluency with faithfulness to the source. Recent large language models (LLMs) often produce fluent translations, but it remains unclear whether fluency corresponds to semantic preservation in literary text. We examine this relationship using 130,486 translated paragraphs from 106 novels in 16 source languages, including human, Google Translate, and TranslateGemma translations. Fluency is measured as original-likeness with a translationese classifier trained on paragraph part-of-speech n-grams, and faithfulness with the automatic translation evaluation metric COMET-KIWI. We control for paragraph length and find a consistent negative correlation between fluency and faithfulness. The pattern appears for both human and Google Translate, but is weaker and often non-significant for TranslateGemma. These results show that segment length matters for automatic evaluation and suggest a tradeoff between fluency and faithfulness in literary translation.