Lexical lectometry in corpus-based translation studies: combining profile-based correspondence analysis and logistic regression modeling

The present study addresses the long-standing issue in corpus-based translation studies that translated texts differ from non-translated texts in the same language, irrespective of text type or source language. We investigate whether this claim is empirically verifiable for a variety of lexical variables in different Dutch varieties or lects (different text types and translated versus non-translated language). By means of profile-based correspondence analysis, linguistic distances are measured and visualized between the lects. Finally, logistic regression modeling enables us to determine the exact impact of the lects on the lexical choices. The results indeed reveal significant differences between translated and non-translated texts, but – contrary to what is generally assumed – these differences are not independent of source language and text type. Keywords: corpus-based translation studies; conservatism, lexical onomasiological variation; correspondence analysis; logistic regression