Validation of Multiple Path Translation for SNOMED CT Localisation.

Area of Science:

• Medical Informatics
• Computational Linguistics
• Natural Language Processing

Background:

• Clinical terminology localization is crucial for global healthcare.
• Existing machine translation tools often struggle with specialized medical terms.
• Human translators require efficient tools to manage terminology translation.

Purpose of the Study:

• To evaluate the Multiple Translation Paths (MTP) approach for clinical terminology localization.
• To assess MTP's effectiveness in generating plausible translation candidates compared to direct methods.
• To validate MTP using SNOMED CT benchmarks and multiple translation engines.

Main Methods:

• The MTP approach was implemented, leveraging web-based machine translation outputs.
• A scored list of translation candidates was generated for input terminology codes.
• Validation was performed using four SNOMED CT benchmarks and three translation engines for German.
• Performance was compared against single best-performing direct-translation approaches.

Main Results:

• MTP demonstrated a significant advantage in generating plausible translation candidate lists for German.
• A moderate advantage was observed for the top-ranked MTP translation candidate over direct methods.
• The study provides initial validation results for the MTP approach in a clinical context.

Conclusions:

• The MTP approach shows promise in supporting clinical terminology localization.
• MTP enhances the efficiency and quality of translation candidate generation for medical terms.
• Further research can explore MTP's application across different languages and terminologies.