A Corpus-based Analysis of Chinese–English Translation Naturalness for Translation Education

Abstract

Naturalness is an important aspect of translation quality that is frequently used to describe both a good translation and a translator’s competence. While an emphasis on naturalness in translation has encouraged translators to pursue it, exactly how it is measured, appraised, or achieved is less clear. This thesis uses corpora to investigate and analyse the naturalness of student Chinese–English translations in comparisons with those of professionals and non-translations in native English, and has the objectives to quantitatively explain and measure variation in translation naturalness, and to identify ways to improve it. Using Toury’s Translation Laws as a theoretical foundation, 34 linguistic features associated with variation in Chinese–English translation naturalness are recognised. A ‘Translation Naturalness Analysis Corpus’ compiled for this research (393,348 English tokens) comprises ‘Student Translation,’ ‘Professional Translation,’ and ‘Non-translation’ subcorpora, populated with 1800 text samples organised into fiction, magazine, and news genres. Unidimensional analyses first identify distinctive patterns in 34 features used to distinguish the naturalness of student translations from those of professional translations and non-translations in the target language. Multidimensional analyses then reveal five dimensions underlying these 34 features. Distributions of the subcorpora along each dimension explain variation in naturalness, and are used to measure the naturalness of student translations against reference lines set by professional translations and non-translations. Multinominal logistic regression analyses confirm the varied effects of dimensions on translation naturalness, which specific suggestions to improve student translation naturalness can be based on. Results imply that translation naturalness and its measurement can be quantified using corpus-based features, that translation naturalness assessment should consider the influences of genre and tolerance to unnaturalness, and that strategies to improve translation naturalness should: be genre-specific; for efficiency, focus on dimensions with greater effects; and be implemented at different training stages. Implications and possible applications of this research are discussed. In conclusion, a model is proposed for future corpus-based translation naturalness studies to follow, which improves research objectivity, generalisability, corpus triangulation, and application of results in an educational context.