A Robust Model for Effective Spam Detection Based on ALBERT

The issue of spam detection has grown increasingly important in order to maintain adequate and safe information transmission as a result of the exponential rise of digital communication channels. The most widely used communication methods today are emails and text messages, and Spam volume increases along with the number of users of email and SMS. Therefore, it becomes imperative to have a reliable mechanism for identifying spam. In this research, we propose an effective method for detecting spam utilizing the pre-trained ALBERT model, which classifies messages and emails according to their context. We used a variety of datasets to train our spam detection model, including the SMS spam collection dataset, the Ling-Spam dataset, and the Twitter Spam Detection Competition dataset from UtkMl. Our experiments demonstrate that, compared to all other alternatives, the suggested approach has the best accuracy results, with 99.28%, 97.75%, and 94.43% on the three datasets, respectively. The model's excellent result on UtkMI's Twitter Spam Detection Competition dataset indicates that it has a good chance of being applied to other related problems.