Rancang Bangun Sistem Penerjemah Multibahasa Daerah Maluku Utara dengan Integrasi GPT Sebagai Pelestarian Budaya Lokal
DOI:
https://doi.org/10.70052/jka.v4i2.1368Keywords:
Regional Language Translator, RBMT, GPT, Multilingual, Digital DictionaryAbstract
Penelitian ini membahas pengembangan sistem penerjemah multibahasa daerah berbasis web yang mengintegrasikan metode Rule-Based Machine Translation (RBMT) dan GPT untuk menerjemahkan Bahasa Indonesia dengan Bahasa Ternate, Makian Dalam, dan Galela menggunakan 5.572 kosakata berbasis kamus digital, dimana RBMT digunakan untuk pencocokan kata berdasarkan aturan dan GPT digunakan untuk memperbaiki struktur kalimat agar lebih natural sesuai konteks, dengan sistem yang dikembangkan menggunakan HTML, CSS, JavaScript, serta integrasi API GPT melalui Vercel, kemudian diuji menggunakan 468 kalimat yang menghasilkan kategori hasil berupa terjemahan sesuai target, berbeda bentuk namun bermakna sama, tidak sesuai akibat keterbatasan kosakata, serta ketidaktepatan karena ambiguitas makna (satu kata yang memiliki banyak arti), sehingga hasil penelitian menunjukkan bahwa kombinasi RBMT dan GPT mampu meningkatkan kualitas terjemahan meskipun masih dipengaruhi oleh ambiguitas kata, kelengkapan kamus, serta penggunaan tanda baca dan pemisahan kata yang dapat mempengaruhi makna dan akurasi hasil terjemahan.
This research discusses the development of a web-based regional multilingual translation system that integrates the Rule-Based Machine Translation (RBMT) method and GPT to translate Indonesian into Ternate, Makian Dalam, and Galela languages using 5,572 vocabulary entries based on a digital dictionary, where RBMT is used for rule-based word matching and GPT is used to improve sentence structure to make the translations more natural and contextually appropriate. The system was developed using HTML, CSS, JavaScript, and GPT API integration through Vercel, and was tested using 468 sentences that produced several categories of results, including translations that matched the target exactly, translations with different sentence forms but the same meaning, inaccurate translations due to limited vocabulary, and inaccuracies caused by semantic ambiguity (a single word having multiple meanings). The results of the study indicate that the combination of RBMT and GPT is capable of improving translation quality, although it is still influenced by word ambiguity, dictionary completeness, and the use of punctuation and word separation, which can affect the meaning and accuracy of the translation results.
References
N. Fokaaya, A. Lestari, M. Taha, F. Febriningsih, K. Bahasa, and P. Maluku, “Bahasa-Bahasa Daerah Bahasa-Bahasa Daerah,” 2014.
A. U. Sulami Sibua, Erviani La Ane Mbongo, Widya Astuti Bilo, Mandaian Fitrianti M, Dewilsta Bunga, “Desiminasi Kamus Praktis Bahasa Melayu Dialek Ternate untuk Wisatawan Lokal dan Internasional,” J. Ilm. Wahana Pendidik., vol. 8, no. July, pp. 1–23, 2022.
D. S. Rainu, Kamus Bahasa Indonesia–Galela. Ternate, Indonesia: Kantor Bahasa Maluku Utara, Kementerian Pendidikan dan Kebudayaan, 2016.
R. M. Hamzah, Gamam Ela-Ela. Jakarta, Indonesia: Kementerian Pendidikan, Kebudayaan, Riset, dan Teknologi, 2024.
I. H. Malan, Ada Pipis Suat Ni Liko Li. Jakarta, Indonesia: Kementerian Pendidikan, Kebudayaan, Riset, dan Teknologi, 2024.
N. Saman, Kaka Madoto. Jakarta, Indonesia: Kementerian Pendidikan, Kebudayaan, Riset, dan Teknologi, 2023.
W. M. Hanif, M. Fikry, F. Yanto, and S. K. Gusti, “Bulletin Of Computer Science Research Implementasi Metode RBMT dalam Penerjemahan Bahasa Indonesia ke Bahasa Makassar,” vol. 6, no. 1, pp. 409–417, 2025, doi: 10.47065/bulletincsr.v6i1.935.
V. Raunak, A. Sharaf, Y. Wang, H. H. Awadalla, and A. Menezes, “Leveraging GPT-4 for Automatic Translation Post-Editing,” Find. Assoc. Comput. Linguist. EMNLP 2023, pp. 12009–12024, 2023, doi: 10.18653/v1/2023.findings-emnlp.804.
X. Song, A. Salcianu, Y. Song, D. Dopson, and D. Zhou, “Fast WordPiece Tokenization,” EMNLP 2021 - 2021 Conf. Empir. Methods Nat. Lang. Process. Proc., no. Section 3, pp. 2089–2103, 2021, doi: 10.18653/v1/2021.emnlp-main.160.
D. Liu and S. Xu, “Comparison of hash table performance with open addressing and closed addressing: An empirical study,” Int. J. Networked Distrib. Comput., vol. 3, no. 1, pp. 60–68, 2015, doi: 10.2991/ijndc.2015.3.1.7.
Y. Wang et al., “Tacotron: Towards end-To-end speech synthesis,” Proc. Annu. Conf. Int. Speech Commun. Assoc. INTERSPEECH, vol. 2017-Augus, pp. 4006–4010, 2017, doi: 10.21437/Interspeech.2017-1452.
J. Keiser and D. Lemire, “On-demand JSON: A better way to parse documents?,” Softw. - Pract. Exp., vol. 54, no. 6, pp. 1074–1086, 2024, doi: 10.1002/spe.3313.
J. M. Bintang, M. F. Ashshidiq, and H. F. Dzakwan, “Penerapan Algoritma String Matching dan Regular Expression pada Aplikasi Kamus Besar Bahasa Indonesia (KBBI),” BIOS J. Teknol. Inf. dan Rekayasa Komput., vol. 4, no. 1, pp. 34–41, 2023, doi: 10.37148/bios.v4i1.57.
B. Ferreira, E. S. de Moura, and A. da Silva, Applying burst-tries for error-tolerant prefix search, vol. 25, no. 4. Springer Netherlands, 2022. doi: 10.1007/s10791-022-09416-9.
H. Yang, “The Application of Game-based Phonics Learning Apps in Lower Elementary Teaching,” J. Educ. Humanit. Soc. Sci., vol. 8, pp. 1939–1946, 2023, doi: 10.54097/ehss.v8i.4618.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Anggiah Salim, Abdul Haris Muhammad , Sakina Sudin (Author)
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
