سامانه همایش های دانشگاه آزاد اسلامی واحد نجف آباد

صفحه اصلی / International Conference on Artificial Intelligence; City, Industry and Health

Enhancing Multilingual Spam Detection Using Machine Learning and Synthetic Data Augmentation

نویسندگان :

Mohamad Hosein Ghojavand¹ Hamid Rastegari²

1- Department of Computer Engineering, Na.C., Islamic Azad University, Najafabad, Iran 2- Department of Computer Engineering, Na.C., Islamic Azad University, Najafabad, Iran

کلمات کلیدی :

Spam Detection،Large Language Models،Machine Learning،Data Augmentation،Text Classification

چکیده :

This study investigates the effectiveness of classical machine learning algorithms for multilingual SMS spam detection, focusing on English and the low-resource language Farsi. We conduct a comprehensive evaluation of 18 classical models on original datasets for both languages. To address the challenge of limited data, particularly for Farsi, we augment the datasets with synthetic SMS messages generated using large language models (LLMs). The performance of the models is compared across three configurations: main (real-world data), synthetic (LLM-generated data), and a combination of both. Results show that synthetic data augmentation significantly improves the performance of several models in Farsi SMS spam detection, leading to enhanced accuracy, precision, recall, and F1-scores. Notably, classifiers trained on the combined Farsi dataset consistently outperform those trained solely on the original data. These findings highlight the potential of LLM-generated synthetic data to enhance the effectiveness of classical models in low-resource language contexts. Moreover, the study underscores the value of data augmentation in multilingual text classification tasks where resources are limited. By comparing performance across languages and dataset types, this research provides insights into cross-linguistic spam detection and demonstrates that classical machine learning remains a viable approach when complemented with modern data augmentation techniques.