Reproduce “Large Engagement Networks for Classifying Coordinated Campaigns and Organic Twitter Trends” on OpenResearch

About this paper

Atul Anand Gopalakrishnan, Jakir Hossain, Tugrulcan Elmas, Ahmet Erdem Sariyuce

Researchers at the University at Buffalo and the University of Edinburgh developed Large Engagement Networks (LEN), a large-scale graph classification benchmark dataset of Turkish Twitter engagement networks. This dataset was created to enable the identification of coordinated campaigns (astroturfing) from organic trends, demonstrating that existing graph machine learning models achieve moderate performance on such challenging, real-world network structures.

Key insights

• Large-scale, real-world social networks like LEN present significant challenges for current graph machine learning models, indicating a need for more robust and scalable algorithms.
• The structural information captured by graph models is crucial for identifying coordinated campaigns, as non-graph based methods performed significantly worse.
• Differentiating between similar themes that vary only in their underlying generation process (e.g., news-based campaigns versus organic news trends) proves particularly difficult for current classification methods.

Summary by alphaXiv.