Student Errors in Translations of Instrumental Texts from Russian to English

This thesis presents the first dedicated study of Russian-English translation errors made by trainee translators who are native speakers of English. The study focuses on the translation of workplace and educational texts with a readily discernible implicit translation brief.

The research addresses three key gaps: the paucity of empirical work quantifying translation error distribution, the absence of any Ru-En student translation corpus by English as L1 speakers, and the shortage of annotation tools preserving segment-to-segment relationships between source and target texts.

Two main tools were developed. MANTRA (MANual TRanslation Annotator) is a corpus assembly tool enabling paired annotation of source and target text segments in a CAT tool-like interface. TRISST (TRanslation ISSue Typology) is a new error classification system achieving broad issue coverage and high granularity while emphasizing semantic analysis, with explicit provisions for analysing semantic roles, temporality, modality, commitment, and scalar-structural relationships.

Data were collected from three postgraduate translation students over one academic year. Eight source texts and 22 target texts were fully annotated, corresponding to 3,374 Russian words and 11,321 English words. 554 issues were identified and rated for seriousness, with 652 issue type tags assigned.

Statistical analysis revealed significant correlations between assessment grades and seriousness-weighted error scores when normalized against assessment means. Two distinct error clusters emerged: referential (various lexical issues, including terminology) and relational (conjunctive, scalar-structural, and temporal-modal issues). Principal component analysis suggests text-type clustering, with workplace texts forming a distinct group separate from educational and encyclopedic materials.

The research confirms conventional pedagogical priorities for instrumental text translation: lexical and terminological precision, syntactic clarity and accuracy, and observance of appropriate norm and usage conventions. It suggests two separable core competencies requiring balanced development: basic lexical competence and conceptual text-world understanding. Strong correlation between error-based scoring and holistic assessment grades indicates these approaches measure fundamentally the same construct.