Social media behavior analyzed under sparse data and noisy signals
Characterizing Social Media Behavior under Limited Observability: Retweet Link Prediction and Multimodal Intention Recognition
Social and Information Networks
Summary
Social media interactions can be hard to study when parts of the data are missing or noisy. The authors focused on two questions: predicting who will retweet whom based on a very limited network and identifying user intentions behind posts with text and images but few labels. They tested different graph and language models and found ways to keep prediction quality high even when data was corrupted or sparse. This helps understand both the connections between users and the reasons why they post in challenging data conditions.
What this means in practice
- •For social media analysts: Improve prediction of retweet links in sparse or noisy social networks using robust graph convolutional models.
- •For content moderation teams: Better identify user intentions from limited text-image social media posts by fine-tuning language models with distribution-aware sampling.
Authors
Yifan Wang, Chenwei Wang
Abstract
Social-media behavior can be observed through both interaction structure and multimodal content, yet each view is often incomplete. We study two complementary tasks on separate datasets: predicting retweet links from an extremely sparse graph without node attributes, and generating user intentions from text-image posts with limited annotations. For link prediction, we use a leakage-controlled transductive protocol to compare GCN and GraphSAGE, classical topology heuristics, and three node-input designs. Across repeated runs, a GCN with compact degree and strength features achieves an average precision of 0.9243 and an MCC of 0.7051. Under joint edge-label corruption, topology-based GCNs retain strong ranking performance, whereas fixed-threshold MCC degrades substantially; adding fixed random node identifiers does not improve overall robustness. For intention recognition, we retain the MIKO framework and evaluate LoRA adaptation of three language models together with distribution-aware sampling based on Pearson correlation, cosine similarity, and KL divergence. Fine-tuning consistently improves lexical-overlap metrics, while distribution-aware sampling produces more uniform performance across intention types. Together, the two studies characterize who interacts and why users post when social-media observations are sparse, noisy, or limited.