Can Publishers Trust AI for Translation?

As AI reshapes publishing, research reminds us that literary translation still needs a human touch.

As artificial intelligence tools become more accessible, authors and publishers are increasingly exploring machine translation to reach global audiences faster and at lower costs. But when the manuscript is a creative work, there is more on the line than accuracy. Recent research on literary translation reveals that while AI can help with speed, it fails to capture the imaginative nuances that make fiction and poetry resonate with readers. For editors and translators, this raises a crucial question: Can technology ever preserve the unique human voice in a story?

THE RESEARCH 

In their study “Creativity in Translation: Machine Translation as a Constraint for Literary Texts,” researchers Ana Guerberof-Arenas and Antonio Toral from the University of Groningen examined how machine translation affects creativity in literary works. Their work focused on measuring creativity when the same source text was translated through three different translation types.

The source text to be translated was a short story by Kurt Vonnegut, which was translated from English into Catalan and Dutch through these approaches: pure human translation (HT), pure machine translation (MT), and human post-editing (PE) of machine output.

Five professional reviewers, blind to the translation type, assessed the texts for creativity via a questionnaire. The researchers observed that “creativity combines acceptability (i.e. something of value, fit for purpose) and novelty (i.e. new, original)” (2022, 6). In addition to asking the reviewers to assess acceptability (through the perceived severity of translation errors and of the rigor of translation successes) and novelty (through the number of creative shifts, or instances when the translation deviates from word-for-word translation to use innovative ways to convey the source text’s meaning), the questionnaire also asked the professionals to rate the overall quality of the translation.

The questionnaire results indicate that all five professionals rated the HT the highest in terms of overall quality. In addition, the HT was perceived to have the lowest severity of errors, the highest rigor of successes, and the highest scores for creative shifts. Even when post-edited by a human, the translations appear to have remained constrained by the machine’s initial linguistic and stylistic choices. Human translators, by contrast, produced more varied and inventive renderings by adapting tone, rhythm, and imagery to preserve the author’s artistic intent.

The researchers stated that “from the data analysed herein, it appears that using MT (by means of PE) hinders the effectiveness of the translation process, because the translator becomes the evaluator and not the creator, and therefore the mechanisms and phases of creativity are not set into motion” (2022, 26).

Ultimately, the study concludes that while MT and PE can make translation faster, they do so at the cost of the literary and aesthetic qualities that define creative writing.

THE IMPLICATIONS

This research suggests that, despite advances in AI, human translators remain essential to literary publishing. Publishers working with more creative works may consider investing in the more costly process of human translators, and leave MT and PE for more technical manuscripts. For editors, translators, and writers worried about being replaced by machines, these findings are reassuring; creativity, nuance, and emotional resonance cannot be automated. The study highlights that literary translation depends on human sensitivity to meaning and to the qualities of a writer’s voice.

As the publishing industry adopts more machine-assisted processes, professionals who understand language and culture deeply will be a breath of fresh air with their ability to understand and edit text on a deeper, human level. Students entering the field may want to invest in learning a second language because the need for human creativity in translation isn’t going anywhere. When it comes to literature, the human touch is essential and irreplaceable. 

To learn more about using AI for literary translation, read the full article: 

Guerberof-Arenas, Ana, and Antonio Toral. “Creativity in Translation: Machine Translation as a Constraint for Literary Texts.” Translation Spaces 11 (2): 184–212. https://doi.org/10.48550/arXiv.2204.05655 

—Ellie Lallatin, Editing Research

FEATURE IMAGE BY 200DEGREES

Find more research

Read Julia Provost’s Editing Research article for another perspective on AI and translation: “The Need for Human Translators in Machine Translation.”

Take a look at Brianne Barrus’s Editing Research article to learn more about translation and accessibility: “Around the World in 80 Edits: The Perception of Multilingual Articles.”