A Corpus of Comparable Student Translations as an Educational Instrument: The Case of Louise Glück’s “Poem”

Abstract

Problem statement. Poetry translation in the university classroom is often discussed through isolated examples, which makes it hard to observe recurring decisions across a cohort. This paper examines 21 Ukrainian student translations produced in 2023 for a university translation contest, all based on a contemporary English-language source text: Louise Glück’s “Poem.” Independent contest submissions are treated as a small parallel corpus that reveals recurring challenges and strategies. Relevance of the Research. A contest corpus lets educators see repeated student choices and respond with targeted feedback. Such corpora are especially informative in Ukrainian translator training because students may recast Anglophone free verse into familiar rhyme or regular meter, shift register, or transfer English structures into Ukrainian phrasing. Aim of the Research. The study aims (1) to describe the qualitative and quantitative profile of the corpus; (2) to systematize recurring decisions in translating an English free-verse lyric into Ukrainian; and (3) to show how an annotated parallel corpus can support form-sensitive training. Research Results. The corpus contains 21 Ukrainian translations of “Poem” (16 lines; 6–4–3–3 stanza pattern). Only 6 translations preserve stanza segmentation; 15 reshape line breaks and/or stanza boundaries; 8 add rhyme or a regular rhythm; and 12 omit the translated title and/or the author’s name. Recurring challenges include lexical ambiguity in image-bearing words, natural noun-phrase and collocational choices, register control, stable deixis and voice (including unnecessary gender marking of the lyrical “I”), and interference from English. Conclusions. A close reading of the contest translations helps identify recurring student difficulties, supports focused classroom practice, and makes assessment more consistent; it also provides students with concrete examples for explaining and defending their solutions. The analysis suggests that form is a meaning-bearing layer: changes in stanza and line breaks, the addition of rhyme or regular rhythm, and omissions of paratext (title/author) tend to coincide with semantic compression, decorative additions, and pragmatic reframing, shifting the text from translation toward adaptation. Research Prospects. Future work will expand the corpus across poems and years, test annotation reliability, and build a searchable classroom resource. Comparing pre-LLM and post-LLM student corpora can track changing difficulty profiles and inform ethically grounded AI-assisted pedagogy.