Technology for Large-Scale Translation of Clinical Practice Guidelines: A Pilot Study of the Performance of a Hybrid Human and Computer-Assisted Approach
Bibliographic record
Abstract
BACKGROUND: The construction of EBMPracticeNet, a national electronic point-of-care information platform in Belgium, began in 2011 to optimize quality of care by promoting evidence-based decision making. The project involved, among other tasks, the translation of 940 EBM Guidelines of Duodecim Medical Publications from English into Dutch and French. Considering the scale of the translation process, it was decided to make use of computer-aided translation performed by certificated translators with limited expertise in medical translation. Our consortium used a hybrid approach, involving a human translator supported by a translation memory (using SDL Trados Studio), terminology recognition (using SDL MultiTerm terminology databases) from medical terminology databases, and support from online machine translation. This resulted in a validated translation memory, which is now in use for the translation of new and updated guidelines. OBJECTIVE: The objective of this experiment was to evaluate the performance of the hybrid human and computer-assisted approach in comparison with translation unsupported by translation memory and terminology recognition. A comparison was also made with the translation efficiency of an expert medical translator. METHODS: We conducted a pilot study in which two sets of 30 new and 30 updated guidelines were randomized to one of three groups. Comparable guidelines were translated (1) by certificated junior translators without medical specialization using the hybrid method, (2) by an experienced medical translator without this support, and (3) by the same junior translators without the support of the validated translation memory. A medical proofreader who was blinded for the translation procedure, evaluated the translated guidelines for acceptability and adequacy. Translation speed was measured by recording translation and post-editing time. The human translation edit rate was calculated as a metric to evaluate the quality of the translation. A further evaluation was made of translation acceptability and adequacy. RESULTS: The average number of words per guideline was 1195 and the mean total translation time was 100.2 minutes/1000 words. No meaningful differences were found in the translation speed for new guidelines. The translation of updated guidelines was 59 minutes/1000 words faster (95% CI 2-115; P=.044) in the computer-aided group. Revisions due to terminology accounted for one third of the overall revisions by the medical proofreader. CONCLUSIONS: Use of the hybrid human and computer-aided translation by a non-expert translator makes the translation of updates of clinical practice guidelines faster and cheaper because of the benefits of translation memory. For the translation of new guidelines, there was no apparent benefit in comparison with the efficiency of translation unsupported by translation memory (whether by an expert or non-expert translator).
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 teacher head, 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".