How AI Can be Used for Governance of Messaging Services: A Study on Spam Classification Leveraging Multi-Channel Convolutional Neural Network

Abstract

Over the past decade, there has been a meteoric evolution in Internet Messaging Services and although these services have become ingrained in our everyday life, SMS service remains an essential form of communication service. The omnipresence of SMS has also given rise to unsolicited and junk messages which has motivated researchers to use machine learning and deep learning to detect such spam messages. Studies using deep learning have shown promising results for spam classification, and in this paper, extending these studies, we have proposed a Multi-Channel CNN architecture with static and dynamic embeddings for SMS spam classification. UCI’s SMS spam collection dataset along with several personally collected text messages are used to create a rich dataset for training the models. The proposed model has an accuracy of 96.12% and overcomes certain disadvantages associated with some of the state-of-the-art deep learning models.