Predicting Information Pathways Across Online Communities

Abstract

The problem of community-level information pathway prediction (CLIPP) aims at predicting the transmission trajectory of content across online communities. A successful solution to CLIPP holds significance as it facilitates the distribution of valuable information to a larger audience and prevents the proliferation of misinformation. Notably, solving CLIPP is non-trivial as inter-community relationships and influence are unknown, information spread is multi-modal, and new content and new communities appear over time. In this work, we address CLIPP by collecting large-scale, multi-modal datasets to examine the diffusion of online YouTube videos on Reddit. We analyze these datasets to construct community influence graphs (CIGs) and develop a novel dynamic graph framework, INPAC (Information Pathway Across Online Communities), which incorporates CIGs to capture the temporal variability and multi-modal nature of video propagation across communities. Experimental results in both warm-start and cold-start scenarios show that INPAC outperforms seven baselines in CLIPP.

arXiv commentIn Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD'23)

References in corpus (31)

• Adam: A Method for Stochastic Optimization
• An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• BPR: Bayesian Personalized Ranking from Implicit Feedback
• Neural Graph Collaborative Filtering
• Session-based Recommendation with Graph Neural Networks
• Predicting Dynamic Embedding Trajectory in Temporal Interaction Networks
• MiniLM: Deep Self-Attention Distillation for Task-Agnostic Compression of Pre-Trained Transformers
• DeepInf: Social Influence Prediction with Deep Learning
• Predict then Propagate: Graph Neural Networks meet Personalized PageRank
• LightGCN: Simplifying and Powering Graph Convolution Network for Recommendation
• Mining Latent Structures for Multimedia Recommendation
• Evaluation of Session-based Recommendation Algorithms
• Community Interaction and Conflict on the Web
• EDITS: Modeling and Mitigating Data Bias for Graph Neural Networks
• Efficiently Leveraging Multi-level User Intent for Session-based Recommendation via Atten-Mixer Network
• Inductive Representation Learning in Temporal Networks via Causal Anonymous Walks
• Continual Learning on Dynamic Graphs via Parameter Isolation
• TwHIN: Embedding the Twitter Heterogeneous Information Network for Personalized Recommendation
• Evolutionary Preference Learning via Graph Nested GRU ODE for Session-based Recommendation
• Predicting Opinion Dynamics via Sociologically-Informed Neural Networks
• Deep Temporal Graph Clustering
• Racism is a Virus: Anti-Asian Hate and Counterspeech in Social Media during the COVID-19 Crisis
• Affective Signals in a Social Media Recommender System
• Towards Fine-Grained Reasoning for Fake News Detection
• Code Recommendation for Open Source Software Developers
• Implicit Session Contexts for Next-Item Recommendations
• SVD-GCN: A Simplified Graph Convolution Paradigm for Recommendation
• A Survey on Incremental Update for Neural Recommender Systems
• TerViT: An Efficient Ternary Vision Transformer
• Whole Page Unbiased Learning to Rank