Slator 2022 Language Industry Market Report
Data and Research, Slator reports
100-page flagship report on market size, buyer-segments, competitive landscape, sales and marketing insights, language tech and more.
The CREAMT project, led by Ana Guerberof Arenas and Antonio Toral, uses a novel, interdisciplinary approach to assess how effective machine translation (MT) is in literary translation. CREAMT was funded by the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant.
The project focuses on the creative aspect of literary texts and the reader’s experience, through the translation of a short story by Kurt Vonnegut from English into Catalan and Dutch.
Articulated in two phases, the first phase analyzes reproductions and creative shifts in three modalities: MT, MT post-editing (PE), and human translation without aid (HT), and in two languages: Catalan and Dutch. The second phase measures the reader’s experience using narrative engagement and enjoyment scales borrowed from psychology, communication, and literary studies.
A paper detailing the results of the project, “Creativity in translation: machine translation as a constraint for literary texts,” was published in mid-April on academic pre-print platform arXiv. Among the hundreds of papers on machine translation published on arXiv, few looked into the topic of creativity.
The aim of the study was to explore creativity (understood to involve novelty and acceptability) from a quantitative perspective. Acceptability was operationalized by the number and severity of errors in the translated texts; and the praises (kudos) and novelty by the number of creative solutions provided for a given problem posed by the source text.
The HT and PE versions were provided by four professional translators specializing in literary translation. A post-assignment questionnaire was also used to gather the translators’ opinions on the quality of MT output.
The reviewers were unanimous in ranking HT as an “extremely good translation,” MT as an “extremely bad translation,” and PE translation as “neither good nor bad.” The results also demonstrated that HT had the highest creativity score, followed by PE and, lastly, MT.
The translators thought that during post-editing, they were “primed or conditioned” by the text they were given, with one participant stating it was like having “a corset,” and that it was “difficult to step out” of the translation that was already there.
Moreover, the translators were unable to set into motion their own mechanisms and depart from the source text, which is one of the conditions for creative translation.
In its current state, machine translation trained on literary data does not have the capability for creative translation, as it renders literal solutions to translation problems, completely omitting creativity.
Although practical, the participants mentioned that MT can only be partially useful in literary translation, particularly for those parts of the text that are of a more straightforward, mechanical nature.
They stated that MT “does not yet have an ‘eye for context’ and ‘no eye for style’ and it misses the coherence of a whole text, only partially working at a sentence level.”At the end of the day, an ideal tool for literary translation would be human-centric. Participants recommended that technology should be in support of the translator rather than merely offering a translation solution.