Leverage knowledge graph and GCN for fine-grained-level clickbait detection.

Area of Science:

• Computer Science
• Information Science
• Social Science

Background:

• Clickbait, deceptive online titles, erodes social trust and hinders information acquisition.
• Existing clickbait detection methods are often binary, using only titles, leading to low performance and limited research utility.
• There is a need for advanced clickbait detection models capable of fine-grained analysis and explainability.

Purpose of the Study:

• To propose a novel clickbait detection model.
• To improve the accuracy and granularity of clickbait detection.
• To provide a foundation for empirical studies on clickbait's societal impact.

Main Methods:

• Developed a novel clickbait detection model integrating a knowledge graph, graph convolutional network, and graph attention network.
• Employed a fine-grained detection approach rather than traditional binary classification.
• Utilized a real-world dataset for experimental evaluation.

Main Results:

• The proposed model significantly outperformed classical and state-of-the-art baselines in clickbait detection.
• Achieved fine-grained clickbait detection capabilities.
• Demonstrated explainability through the graph attention network component.

Conclusions:

• The novel model offers a more effective and explainable solution for clickbait detection.
• Fine-grained analysis provides a basis for future empirical research on clickbait.
• This work represents the first integration of knowledge graphs and deep learning for explainable clickbait detection.