WSDM 2022
182 papers
The 2022 edition of WSDM featured 279 papers centered on advancements in information retrieval and data mining. A primary focus was the refinement of recommendation systems, with research addressing sequential recommendation, cold-start personalization, and the mitigation of representation collapse. Contrastive learning and intent-preserving data augmentation emerged as key methodologies for improving these systems.
The program also emphasized robust evaluation and learning frameworks, including counterfactual evaluation of decision-making policies and doubly robust off-policy evaluation for ranking. Other research directions included unbiased learning to rank from biased user interactions and the application of variational autoencoders for collaborative filtering. Technical explorations extended to graph-based learning, relational triple extraction, and the analysis of co-evolving sequences. Prolific contributors to this year's proceedings included Suhang Wang, Ruocheng Guo, and Leyu Lin. Additionally, the conference addressed social challenges through tutorials on combating online hate speech by examining content, psychology, and user behavior.
Top Research Directions
Research directions most represented by WSDM 2022 papers.
Top Authors
Impactful Papers
Papers from WSDM 2022 that shaped research clusters.
Foundational
PipAttack: Poisoning Federated Recommender Systems for Manipulating Item Promotion
foundational
Securing Federated Learning Against Mode
Influential
Learning Multi-granularity Consecutive User Intent Unit for Session-based Recommendation
influential
Balancing Model Complexity and Computati
Evaluating Mixed-initiative Conversational Search Systems via User Simulation
influential
Interactive Evaluation of Conversational
DualDE: Dually Distilling Knowledge Graph Embedding for Faster and Cheaper Reasoning
influential
Scaling Graph Embeddings Beyond the Memo
PipAttack: Poisoning Federated Recommender Systems for Manipulating Item Promotion
influential
Data Poisoning and Shilling in Recommend
EvoKG: Jointly Modeling Event Time and Network Structure for Reasoning over Temporal Knowledge Graph
influential
Embedding Temporal Dynamics in Knowledge
Finding a Concise, Precise, and Exhaustive Set of Near Bi-Cliques in Dynamic Graphs
influential
Deconstructing Complex Models Through Sy