An automated study of translationese indicators in scientific paper translations

Authors:
Abstract:

The rise in global scientific publication output and international knowledge exchange has heightened the demands for the quality of scientific paper translations. Among the factors influencing their quality, the phenomenon of translationese stands out due to its characteristics such as simplification, explicitation, normalization, and interference. Not only do they differentiate translations from non-translated texts, but they might also be indicative of translation quality, suggesting the need to consider them when evaluating the quality of a translation and deciding on its further post-editing. However, calculating and analyzing the indicators of these characteristics is a lengthy and labor-intensive process, which creates the need for tools to automate it. The article offers a solution to this issue through the use of a specialized tool designed for compiling text corpora, identifying translationese indicators in individual texts and entire corpora, as well as comparing collected data across corpora. A comparable corpus of sociology-related scientific papers is used as material for validating the accuracy of the tool – this corpus includes two textual subcorpora of the same style, genre, subject matter, and size but of different origin: the first subcorpus comprised original (non-translated) Russian texts, while the second one consisted of English-to-Russian translations. The results of the analysis reveal the presence of such translationese indicators in translations as reduced lexical variety, lexical uniqueness, and lexical repetition; increased text explicitness characterized by the frequent use of personal, possessive, and demonstrative pronouns and discourse markers used for clarifying information and contrast; a more frequent use of direct word order and preference for verbs rather than their substantivized counterparts; the tendency to use relative clauses more often than participial phrases. The article discusses the implications of these findings and outlines possible strategies for reducing their impact on English-to-Russian scientific paper translations.