Enhancing Clickbait Detection with Cross-Modal Topic Modeling in Social Networks

Clickbait content poses a significant challenge in on-line information consumption, affecting user engagement, trust, and overall content quality. This paper presents a novel approach for clickbait detection leveraging topic modeling techniques across different modals of data. We propose a methodology where different topic modeling algorithms are employed to extract latent topics present in the text. Through this process, the underlying thematic structures within the content are revealed. Topic modeling algorithms like Latent Dirichlet Allocation (LDA), Latent Semantic Analysis(LSA) and BERTopic were utilized to filter out clickbait content from a diverse corpus of web content. The results show that different topic modeling techniques exhibit varying degrees of success in clickbait detection, with some methods excelling in certain scenarios with a recall of 90%, while others offer a more generalized approach. Experiments were performed on real-world datasets and the findings of this study highlight the effectiveness of the proposed method in identifying clickbait, thereby providing a valuable tool for content moderation and enhancing user experience.