Full text loading...
and Stella Neumann1
Abstract
How does the variability of the language system affect translational language use? Despite inconclusive results of earlier corpus studies, machine learning approaches reach high accuracies in distinguishing translated from non-translated texts. Translations must therefore involve linguistic patterns easy to spot by computers, but harder to spot by the human analyst. Consequently, the differences have been characterized as small, but systematic. This article adopts a quantitative corpus-based approach to examining shifts in probability between translated and non-translated language. To this end, we investigate how translators handle non-agentive constructions, a phenomenon displaying contrastive usage differences in English and German. A multifactorial analysis of 7441 instances shows that differences are indeed small and that they stem from a multitude of factors. Accounting for these factors turns out to be challenging even for a sophisticated statistical procedure. Therefore, the case of how subtle the effects of translation are cannot be settled just yet.
Article metrics loading...
Full text loading...
References
Data & Media loading...