Quality Evaluation of Roche’s NMT system DIAlingo

Diplôme Master Quality Evaluation of Roche’s NMT system DIAlingo eng 2023 author Racaj Adelina 941411 director Gerlach Johanna 0000-0002-3371-4021 645950 Département de traitement informatique multilingue (TINFM) director Krämer Vivien Neural machine translation (NMT) brings new quality standards in machine translation, and new ways of working for both Language Service Providers as well as multinational companies. Since quality is the major determinant of success, companies need to ensure that their NMT models perform at the required quality standards by improving their NMT model's performance constantly. To do this, the NMT model's performance needs to be monitored. Traditionally, this entails human evaluation performed by post-editors who score the machine translation output according to specific evaluation criteria. Recently, computer linguists have proposed automatic evaluation metrics. This thesis empirically analyses 6 different automatic evaluation metrics to investigate which automatic metric correlates best with human evaluation for the English-German model, which is part of Roche Diagnostics' NMT system DIAlingo, and how the inclusion of NMT impacts the efficiency of translation processes within Roche Diagnostics International Ldt. Neural Machine Translation Google AutoML Translation Human Evaluation Automatic Evaluation Quality Evaluation Metrics Dewey 410.2 Maîtrise universitaire en traitement informatique multilingue 112 Département de traitement informatique multilingue 92 false f5a52c3b-125d-479f-8b5b-cd8b679f1d60 Master thesis Private Master_Thesis_MATIM_RACAJ.pdf 19911208 application/pdf 6feca4be65a3ffb97f5ac5222046d51787fb06adfaa573ec796e49db5df13fc7 closedAccess Faculté de traduction et d'interprétation / Département de traitement informatique multilingue

Title	Published in	Access level	OA Policy	Year	Views	Downloads

Quality Evaluation of Roche’s NMT system DIAlingo				2023	96	0

Archive ouverte UNIGE

Racaj, Adelina