An Attention-Based Neural Network Using Human Semantic Knowledge and Its Application to Clickbait Detection

Clickbait is a commonly used social engineering technique to carry out phishing attacks, illegitimate marketing, and dissemination of disinformation. As a result, clickbait detection has become a popular research topic in recent years due to the prevalence of clickbait on the web and social media. In this article, we propose a novel attention-based neural network for the task of clickbait detection. To the best of our knowledge, our work is the first that incorporates human semantic knowledge into an artificial neural network, and uses linguistic knowledge graphs to guide attention mechanisms for the clickbait detection task. Extensive experimental results show that the proposed model outperforms existing state-of-the-art clickbait classifiers, even when training data is limited. The proposed model also performs better or comparably to powerful pretrained models, namely, BERT, RoBERTa, and XLNet, while being much more lightweight. Furthermore, we conducted experiments to demonstrate that the use of human semantic knowledge can significantly enhance the performance of pretrained models in the semisupervised domain such as BERT, RoBERTa, and XLNet.