The Translation of Multilingual Signboards in Mataram City Using Google NMT

This research investigates the use of Google Neural Machine Translation (GNMT) in translating multilingual signboards across Mataram City. It focuses on examining the accuracy and readability of GNMT translations to assess their effectiveness in representing real-world public communication. Based on the formulated research questions, this study aims to analyze the accuracy and readability of Google Neural Machine Translation (GNMT) in translating multilingual signboards in Mataram City and to assess the translation quality of Google NMT in comparison to human translation on public multilingual signboards. The research was conducted from June 9 to July 4, 2025, along Jl. Majapahit, Mataram City, West Nusa Tenggara. The population of this study encompasses all multilingual signboards found in public spaces across Mataram City, while the sample was drawn specifically from multilingual signboards located along Jl. Majapahit, Mataram. A total of 209 signboards, including both monolingual and multilingual types, were purposively selected based on readability, completeness, and accessibility for analysis. The study found that GNMT achieved an average accuracy and readability score of 4.37 compared to the human score of 5.0, performing well on clear and structured texts but showing limitations in translating cultural and idiomatic expressions, which require human refinement for contextual accuracy. In conclusion, GNMT effectively translates multilingual signboards with 100% OCR success, performing well on clear texts but struggling with idiomatic and culturally nuanced expressions, thus highlighting the need for human involvement to ensure contextual and cultural accuracy.