📅 31 August 2025
DOI: 10.35671/telematika.v18i2.3140
Automatic Analysis of Natural Disaster Messages on Social Media Using IndoBERT and Multilingual BERT
Safitri; Yasmin Dwi; Lambung Mangkurat University; Faisal; Mohammad Reza; Lambung Mangkurat University; Kartini; Dwi; Lambung Mangkurat University; Saragih; Triando Hamonangan; Lambung Mangkurat University; Abadi; Friska; Lambung Mangkurat University; Bachtiar; Adam Mukharil; Japan Advanced Institute of Science and Technology
Telematika
Universitas Amikom Purwokerto
📄 Abstract
Information about natural disasters disseminated through social media can serve as an important data source for mitigation processes and early warning systems. Social media platforms, such as X (formerly known as Twitter), have become primary channels for conveying real-time information, especially during disaster emergencies. With the large amount of unstructured disaster-related text that must be processed, the main challenge is accurately filtering and classifying messages into three categories: eyewitness, non-eyewitness, and don’t know. This research aims to compare the performance of four BERT-based natural language processing models, namely IndoBERT, IndoBERT with Masked Language Modeling (MLM), Multilingual BERT, and Multilingual BERT with MLM, in classifying Indonesian-language disaster messages. The dataset used in this study was obtained from previous research and publicly available data on GitHub, consisting of annotated messages related to floods, earthquakes, and forest fires. The method applied is a deep learning approach using the hold-out technique with an 80:20 ratio for training and testing data, and the same ratio applied to split the training data into training and validation subsets, with stratification to maintain balanced class proportions. In addition, variations in batch size were explored to evaluate their effect on model performance stability. The results show that the IndoBERT model achieved the highest performance on the flood and earthquake datasets, with accuracies of 80.67% and 81.50%, respectively. Meanwhile, IndoBERT with MLM pre-training recorded the highest accuracy on the forest fire dataset, 88.33%. Overall, IndoBERT demonstrated the most consistent and superior performance across datasets compared to the other models. These findings indicate that IndoBERT has strong capabilities in understanding Indonesian disaster-related text, and the results can be used as a foundation for developing automatic classification systems to support real-time disaster monitoring and early warning applications
🔖 Keywords
#Deep Learning; Social Media; Natural Disaster; IndoBERT; Multilingual BERT
ℹ️ Informasi Publikasi
Tanggal Publikasi
31 August 2025
Volume / Nomor / Tahun
Volume 18, Nomor 2, Tahun 2025
📝 HOW TO CITE
Safitri, Yasmin Dwi; Lambung Mangkurat University; Faisal, Mohammad Reza; Lambung Mangkurat University; Kartini, Dwi; Lambung Mangkurat University; Saragih, Triando Hamonangan; Lambung Mangkurat University; Abadi, Friska; Lambung Mangkurat University; Bachtiar, Adam Mukharil; Japan Advanced Institute of Science and Technology; , "Automatic Analysis of Natural Disaster Messages on Social Media Using IndoBERT and Multilingual BERT," Telematika, vol. 18, no. 2, Aug. 2025.
ACM
ACS
APA
ABNT
Chicago
Harvard
IEEE
MLA
Turabian
Vancouver
Download Citation
0
Sitasi
Cek Google Scholar