Clinically-weighted terms in the evaluation of MeSH translations: BabelMeSH in Portuguese as a model.

We propose a new method of evaluating the accuracy of machine translation by clinical weighting. Weight is determined by its importance in the clinical diagnosis. A new algorithm was developed based on clinical term weighting. We compared simple word alignment, clinical weighting algorithm and human review by translation efficiency. The results were 65%, 86% and 90% respectively. This method encourages a quantitative evaluation of translation accuracy and facilitates automation. Clinical weighting promotes better evaluation.