Clickbait Detection for Amharic Language using Deep Learning Techniques

Abstract

Because of, the increasing number of Ethiopians who actively engaging with the Internet and social media platforms, the incidence of clickbait is becomes a significant concern. Clickbait, often utilizing enticing titles to tempt users into clicking, has become rampant for various reasons, including advertising and revenue generation. However, the Amharic language, spoken by a large population, lacks sufficient NLP resources for addressing this issue. In this study, the authors developed a machine learning model for detecting and classifying clickbait titles in Amharic Language. To facilitate this, authors prepared the first Amharic clickbait dataset. 53,227 social media posts from well-known sites including Facebook, Twitter, and YouTube are included in the dataset. To assess the impact of conventional machine learning methods like Random Forest (RF), Logistic Regression (LR), and Support Vector Machines (SVM) with TF-IDF and N-gram feature extraction approaches, the authors set up a baseline. Subsequently, the authors investigated the efficacy of two word embedding techniques, word2vec and fastText, with Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU) deep learning algorithms. At 94.27% accuracy and 94.24% F1 score measure, the CNN model with the rapid Text word embedding performs the best compared to the other models, according to the testing data. The study advances natural language processing on low-resource languages and offers insightful advice on how to counter clickbait content in Amharic.