Towards a machine learning approach to the analysis of indirect translation

Abstract

Despite its importance in a globalized world, indirect translation is a peripheral and under-researched topic in translation studies. Existing research on indirect translation is almost exclusively limited to literary translation and focuses mainly on historical aspects. From a methodological perspective, textual analysis based on close reading is the main source of insight into indirect translation, while distant reading using computational approaches remains unexplored. In order to promote methodological innovation, this study gives a replicable demonstration of how to apply supervised machine learning to corpora of indirect translations. The study is based on comparable corpora of proceedings from the European Parliament. Open-access data is used to ensure the replicability of the proposed methodology. Based on the computational findings, the methodological caveats of this approach are discussed.