Deep Dive into Clickbait Secrets: Integrating Multi-modal Features and Leveraging Deep Learning Architectures

Clickbait is a term often used to define articles, web pages, or online content of any type which entices a user to pay attention to it. It is a genre of content which is known for its extreme sensationalist nature and encompasses meticulously crafted content, often in the form of attention-grabbing headlines and captivating thumbnails. Its primary aim is to capture user attention and compel them to click on a link. This allure is achieved through elements that pique curiosity, or exaggerated claims, all with the central objective of driving user engagement, increasing website traffic, and augmenting advertising revenue. However, a common pitfall of clickbait is its frequent failure to meet user expectations, resulting in dissatisfaction. Building up on prior research with the Clustered Clickbait Dataset, this study delves deeper into the domain of deep learning architectures and machine learning algorithms to model and understand this dataset. The study explores multiple deep learning architectures in search of the best-fitting model. Three models are introduced: a text-only model with 98.25% accuracy, an image-only model with 96.5% accuracy, and a multi-modal model with an impressive 98.7% accuracy.