The MTP (multiple translation paths) approach supports human translators in clinical terminology localization. It exploits the results of web-based machine translation tools and generates, for a chosen target language, a scored output of translation candidates for each input terminology code. We present first results of a validation, using four SNOMED CT benchmarks and three translation engines. For German as target language, there was a significant advantage of MTP as a generator of plausible translation candidate lists, and a moderate advantage of the top-ranked MTP translation candidate over single best performing direct-translation approaches.
Keywords: Machine Translation; NLP; SNOMED CT.