Classifying extortion messages using machine learning models and large language model
DOI:
https://doi.org/10.15649/2346030X.5590Keywords:
natural language processing, classifying, machine learning, large language modelAbstract
This paper addresses the problem of classifying extortion messages. The extortion messages contain words that express a threat to someone and are written in Spanish with numerous spelling mistakes and abbreviated words. The concepts of natural language processing, machine learning, and deep learning were applied. A dataset with both extortion and no extorsion messages was constructed; three models were trained and validated: one with a random forest and another with a support vector machine; these models were configured with several hyper parameters for a better analysis. These models are considered traditional models and belong to the machine learning framework. A final model of deep learning was trained with BETO, a spanish BERT release. The results obtained from all the models were evaluated by comparing them with the confusion matrix indicators for each model with several settings. It can be concluded that the models achieve very high accuracy. However, BETO training was time-consuming and expensive, so traditional models perform well on this problem and can be used in similar classification problems, such as the classification of other crimes such as fraud, harassment, and sentiment analysis in specific contexts.
References
[1] K. Guan et al., “IEEE Access Special Section Editorial: 5G and beyond mobile wireless communications enabling intelligent mobility,” IEEE Access, vol. 8, pp. 208892–208897, 2020, doi: 10.1109/ACCESS.2020.3037635.
[2] V. dos Santos Riedo and S. A. Carro, “Um modelo para extração, estruturação, indexação e recuperação de casos clínicos publicados na web,” Colloquium Exactarum, vol. 11, no. 2, pp. 91–100, 2019, doi: 10.5747/ce.2019.v11.n2.e278.
[3] E. Elsaeed, O. Ouda, M. M. Elmogy, A. Atwan, and E. El-Daydamony, “Detecting fake news in social media using voting classifier,” IEEE Access, vol. 9, pp. 161909–161925, 2021, doi: 10.1109/ACCESS.2021.3132022.
[4] M. Callejas Cuervo and M. A. Vélez Guerrero, “Predicción electoral usando un modelo híbrido basado en análisis sentimental y seguimiento a encuestas: elecciones presidenciales de Colombia,” Revista Politécnica, vol. 15, no. 30, pp. 94–104, 2019, doi: 10.33571/rpolitec.v15n30a9.
[5] R. Anggrainingsih, G. M. Hassan, and A. Datta, “CE-BERT: Concise and efficient BERT-based model for detecting rumors on Twitter,” IEEE Access, vol. 11, pp. 80207–80217, 2023, doi: 10.1109/ACCESS.2023.3299858.
[6] S. H. Al Sulaimani and A. Starkey, “Short text classification using contextual analysis,” IEEE Access, vol. 9, pp. 149619–149629, 2021, doi: 10.1109/ACCESS.2021.3125768.
[7] Z. E. Rasjid and R. Setiawan, “Performance comparison and optimization of text document classification using k-NN and Naïve Bayes classification techniques,” Procedia Comput. Sci., vol. 116, pp. 107–112, 2017, doi: 10.1016/j.procs.2017.10.017.
[8] R. Ahamad and K. N. Mishra, “Exploring sentiment analysis in handwritten and E-text documents using advanced machine learning techniques: A novel approach,” J. Big Data, vol. 12, no. 1, Art. no. 11, 2025, doi: 10.1186/s40537-025-01064-2.
[9] J. R. Davalos-Miranda, “Análisis del Sentimiento: Exploración de League of Legends y su comunidad en Twitter,” M.S. thesis, Universidad Internacional de La Rioja (UNIR), 2022.
[10] C. Simón-Gallego, “Comparativa de modelos de aprendizaje profundo para la detección de odio en castellano en medios de información social,” M.S. thesis, Universidad Internacional de La Rioja (UNIR), 2023.
[11] K. Taneja, J. Vashishtha, and S. Ratnoo, “Fraud-BERT: transformer based context aware online recruitment fraud detection,” Discover Computing, vol. 28, Art. no. 9, 2025, doi: 10.1007/s10791-025-09502-8.
[12] I. Belzunegui-Gabilondo, “Creación de un clasificador de sentimientos de alumnos mediante minería de opinión,” M.S. thesis, Universidad Internacional de La Rioja (UNIR), 2023.
[13] M. E. Mswahili, J. Hwang, J. C. Rajapakse, K. Jo, and Y.-S. Jeong, “Positional embeddings and zero-shot learning using BERT for molecular-property prediction,” J. Cheminform., vol. 17, no. 1, Art. no. 17, 2025, doi: 10.1186/s13321-025-00959-9.
[14] Y. Yang and X. Peng, “BERT-based network for intrusion detection system,” EURASIP J. Inf. Secur., vol. 2025, no. 1, Art. no. 11, 2025, doi: 10.1186/s13635-025-00191-w.
[15] N. M. Gardazi, A. Daud, M. K. Malik, A. Bukhari, T. Alsahfi, and B. Alshemaimri, “BERT applications in natural language processing: a review,” Artif. Intell. Rev., vol. 58, no. 6, Art. no. 166, 2025, doi: 10.1007/s10462-025-11162-5.
[16] M. Maktabdar Oghaz, L. Babu Saheer, K. Dhame, and G. Singaram, “Detection and classification of ChatGPT-generated content using deep transformer models,” Front. Artif. Intell., vol. 8, Art. no. 1458707, 2025, doi: 10.3389/frai.2025.1458707.
[17] C. Samper-Quinto, “Análisis de sentimiento en procesamiento cuántico del lenguaje natural en un contexto financiero,” M.S. thesis, Universidad Internacional de La Rioja (UNIR), 2023.
[18] O. Elizarraras and J. Chakraborty, “The use of social media in generating health information support for Spanish users,” Procedia Comput. Sci., vol. 256, pp. 1183–1191, 2025, doi: 10.1016/j.procs.2025.02.227.
[19] K. C. Argomedo Pflücker and A. T. Cóndor Ruiz, “Diseño de un sistema inteligente utilizando procesamiento de lenguaje natural (PLN) para el reconocimiento de mensajes extorsivos,” B.S. thesis, Universidad Nacional de Trujillo, Trujillo, Peru, 2015.
[20] C. M. Liapis, A. Karanikola, and S. Kotsiantis, “Enhancing sentiment analysis with distributional emotion embeddings,” Neurocomputing, vol. 634, Art. no. 129822, 2025, doi: 10.1016/j.neucom.2025.129822.
License
Copyright (c) 2026 AiBi Journal of Research, Administration and Engineering
This work is licensed under a Creative Commons Attribution 4.0 International License.
The journal offers open access under a Creative Commons Attibution License
This work is under license Creative Commons Attribution (CC BY 4.0).
Downloads
