Translating literature is a special challenge. Each sentence of a literary text calls for creativity, imagination, committed decision-making and instincts you can trust. How good is machine translation at negotiating this creative process? Researchers from the University of Surrey and the University of Groningen set out to discover whether digital resources have what it takes to match the skills of literary translators.
With and without machine assistance
The study focused on 2BR02B, a short story written by US literary giant Kurt Vonnegut. A number of translators were given the task of translating this text into Catalan or Dutch using one of three approaches: relying entirely on a computer program (machine translation), translating it themselves with a machine translation as their basis (postediting), or translating without the aid of digital shortcuts (human translation). A panel of independent reviewers then assessed the quality of the translations: which approach resulted in the most readable text and offered the most creative solutions?
The result: human 1, machine 0
Perhaps unsurprisingly, the translations produced by human translators with no help from machines came out on top. Machine translations emerged as the least creative; the computer programs were prone to error and their solutions to tricky translation issues were often too literal. Another finding did turn up a surprise: the results of postediting – machine translations spruced up by human expertise – also proved disappointing. Working on the basis of a rough translation provided by a machine would appear to inhibit a translator’s creativity.
Creative translation calls for problem-solving skills
The study makes it clear that creative translation involves so much more than converting words from one language to another. Staying true to the original work while creating a new reading experience for readers in another language is an intricate balancing act. In many cases, literary translators need to step back from the source text in order to maintain a natural, credible and satisfyingly fluid style. At the same time, they need to find ways of reflecting the original author’s style and the ‘foreign’ context of the story. It’s a process that calls on powers of association, instinct and emotional sensitivity – areas in which machines fall short.
Human-machine collaboration: not always a recipe for success
The study also assessed translations that were postedited: translated by a computer program, then edited by a translator. This sounds like a winning combination, but the results left a lot to be desired. The translators observed that the existing computer translation often interfered with their own thought process. Meanwhile, the panel of reviewers assessed the results as too close to the source text and demonstrating less creativity.
Technology as a smart assistant
Despite these shortcomings, the study also showed that technology can provide worthwhile support in the translation process. The participants noted that digital translation tools can aid consistency, offer fast-track solutions to standard translation problems, and in some cases be a source of valuable inspiration. Literary translators are certainly open to using translation tools as smart assistants, a supporting player that streamlines the translation process without curtailing the translator’s craftsmanship and creative flow.
Find out more about the study: Creativity in translation: Machine translation as a constraint for literary texts. Ana Guerberof-Arenas (University of Surrey, University of Groningen) and Antonio Toral (University of Groningen).
Looking for a creative translation?
Find out more by visiting this page, or contact us at vertaalbureau@taalcentrum-vu.nl
