Performance of machine translators in translating French medical research abstracts to English: A comparative study of DeepL, Google Translate, and CUBBITT
Bibliographic record
Abstract
BACKGROUND: Non-English speaking researchers may find it difficult to write articles in English and may be tempted to use machine translators (MTs) to facilitate their task. We compared the performance of DeepL, Google Translate, and CUBBITT for the translation of abstracts from French to English. METHODS: We selected ten abstracts published in 2021 in two high-impact bilingual medical journals (CMAJ and Canadian Family Physician) and used nine metrics of Recall-Oriented Understudy for Gisting Evaluation (ROUGE-1 recall/precision/F1-score, ROUGE-2 recall/precision/F1-score, and ROUGE-L recall/precision/F1-score) to evaluate the accuracy of the translation (scores ranging from zero to one [= maximum]). We also used the fluency score assigned by ten raters to evaluate the stylistic quality of the translation (ranging from ten [= incomprehensible] to fifty [= flawless English]). We used Kruskal-Wallis tests to compare the medians between the three MTs. For the human evaluation, we also examined the original English text. RESULTS: Differences in medians were not statistically significant for the nine metrics of ROUGE (medians: min-max = 0.5246-0.7392 for DeepL, 0.4634-0.7200 for Google Translate, 0.4815-0.7316 for CUBBITT, all p-values > 0.10). For the human evaluation, CUBBITT tended to score higher than DeepL, Google Translate, and the original English text (median = 43 for CUBBITT, vs. 39, 38, and 40, respectively, p-value = 0.003). CONCLUSION: The three MTs performed similarly when tested with ROUGE, but CUBBITT was slightly better than the other two using human evaluation. Although we only included abstracts and did not evaluate the time required for post-editing, we believe that French-speaking researchers could use DeepL, Google Translate, or CUBBITT when writing articles in English.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.015 | 0.068 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.005 | 0.004 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.004 | 0.004 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.002 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.003 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".