<![CDATA[Artificial Intelligence (AI) has become essential in digital navigation systems such as Google Maps, whose Indonesian Text-to-Speech (TTS) feature still struggles with mispronouncing abbreviations, acronyms, and proper names, often producing speech that sounds unnatural or confusing. This study aims to identify the linguistic and technical causes of pronunciation errors in Google Maps’ Indonesian TTS system. This study employs a qualitative approach combining content analysis and computational linguistics. Audio samples from Google Maps were analyzed to identify recurring pronunciation issues, followed by a technical examination of phonetic modeling, text-to-phoneme conversion, and text-normalization processes within the TTS system. The results show that inaccuracies arise from three key factors: insufficient contextual recognition of abbreviations, limited lexical databases, and weaknesses in text-normalization mechanisms. These findings suggest that the challenges are not purely technical but also linguistic and sociocultural, reflecting the complexity of Indonesian with its diverse dialects, orthography, and local naming practices. The study recommends improving lexical coverage, developing more context-aware linguistic models, and training AI with voice and text data that better represent Indonesian language variation. Overall, this research contributes to the development of a more accurate, natural, and culturally attuned TTS system while underscoring the importance of integrating AI innovation with linguistic sensitivity.]]>
Copyrights © 2025
