Translation Quality in Google Translate: A Syntactic Complexity Analysis of Indonesian–English Output

Abstract

Machine translation has become increasingly prevalent in Indonesia, yet the effect of sentence complexity on translation quality remains underexplored in the context of the Indonesian–English language pair. This study investigates the accuracy, readability, and acceptability of Google Translate outputs across four sentence types (simple, compound, complex, and compound-complex) using Nababan's (2012) Translation Quality Assessment (TQA) framework. A descriptive qualitative method was employed, with 100 sentences purposively selected from the novel Teruslah Bodoh Jangan Pintar and evaluated by two expert raters. Results show that translation quality declines consistently as syntactic complexity increases: simple sentences achieved the highest scores across all three dimensions, while compound-complex sentences performed the lowest. Importantly, readability consistently exceeded accuracy across all sentence types, indicating that Google Translate tends to produce fluent-sounding output that nonetheless contains meaning distortions and unnatural lexical choices. These findings highlight the limitations of NMT systems in processing hierarchical clause relationships, subordination, and contextual nuance, and reinforce the need for human post-editing in complex translation tasks. The study provides a replicable syntactic classification model for future NMT quality research and offers practical guidance for language educators and translation practitioners in Indonesia.