https://ejournal.seaninstitute.or.id/index.php/visualization <p align="JUSTIFY">The "Journal of Data Science" is a real journal that focuses on the field of data science. It covers a wide range of topics related to data analysis, machine learning, statistics, data mining, and related areas. The journal aims to publish high-quality research papers, reviews, and technical notes that contribute to the advancement of data science.</p> <p align="JUSTIFY">The Journal of Data Science welcomes submissions from researchers, academics, and practitioners working in the field of data science. It provides a platform for sharing novel research findings, methodologies, algorithms, and applications in various domains.</p> Sean Institute en-US Journal Of Data Science 3025-2792 https://ejournal.seaninstitute.or.id/index.php/visualization/article/view/8331 <p>The expansion of social media platforms has significantly reshaped marketing communication strategies, particularly in countries with highly active digital audiences such as Indonesia. Among these platforms, TikTok has become a dominant channel for promotional messaging, where short textual captions play a crucial role in attracting user attention. Within this context, the phonological structure of language may influence how promotional messages are perceived, remembered, and shared. This study investigates the phonological characteristics of Indonesian marketing captions on social media using a Natural Language Processing (NLP) framework. A dataset consisting of user-generated promotional captions from TikTok was collected and processed through several stages, including text normalization, tokenization, and noise removal. The cleaned text was then transformed into phoneme sequences using a hybrid grapheme-to-phoneme (G2P) conversion approach. From these phoneme representations, several phonological features were extracted, including vowel–consonant ratios, syllable length distribution, phoneme entropy, and phonotactic surprisal. The analysis reveals several recurring sound patterns in Indonesian promotional language. In particular, captions tend to favor open syllable structures, vowel-rich word formations, and repetitive phoneme sequences that contribute to rhythmic flow and memorability. Words containing dominant vowels such as /a/ and /i/ appear frequently in marketing expressions, suggesting that phonological aesthetics may play a role in shaping audience engagement. Overall, the findings suggest that phonological structure is not merely a linguistic artifact but may also function as a subtle persuasive mechanism in digital marketing discourse. By combining computational phonology with marketing analysis, this study offers new insights into how sound patterns embedded in written promotional texts can influence communication effectiveness in the Indonesian social media landscape.</p> Fitrah Rumaisa Zakiah Darajat Yan Puspitarani Copyright (c) 2026 Journal Of Data Science 2026-05-04 2026-05-04 4 01 1 13