Call for Papers

IDEA invites research on raw (atomic), semantic, and hybrid item identifiers as learned features or token sequences in Web systems. We welcome work on raw-ID embeddings, semantic codes, and hybrid representations for search, recommendation, computational advertising, and related applications.

Our focus spans the lifecycle of these representations: their construction and learning, the evidence supporting their use, and their adaptation as catalogs, interaction signals, and models change. Contributions may study any one ID family or compare multiple designs; a comparison with semantic IDs is not required. We welcome completed methods, empirical findings, deployment experiences, and early work that can benefit from community discussion.

Topics of Interest

We focus on three areas: identifier design and learning, evaluation and comparison of ID representations, and deployment and adaptation. Topics include, but are not limited to, the following. Other applications and explorations of ID-based features and token representations in Web systems are also welcome.

Identifier Design and Learning

  • Raw-ID embeddings, item-specific memorization, and integration with content and behavioral features.
  • Semantic ID construction from content, collaborative signals, and multimodal representations.
  • Discrete representation learning, quantization, tokenization, and codebook design.
  • Hybrid identifiers combining atomic and structured components.
  • Joint learning of ID representations with retrieval, ranking, or generative models.
  • Shared identifier spaces and representation transfer across tasks and domains.

Evaluation and Comparison

  • Evaluation protocols and benchmarks for ID representations, with controlled comparisons that isolate identifier choices from changes in features, models, and training.
  • Memorization and generalization properties of raw-ID embeddings, semantic codes, and hybrid representations, including cold-start and long-tail behavior.
  • Representation quality and downstream utility, including code utilization, collisions, semantic–behavioral alignment, and their relationship to retrieval and ranking effectiveness.
  • Beyond-accuracy effects of identifier design, including diversity, novelty, coverage, fairness, and user utility.
  • Robustness and stability of ID representations under catalog changes, sparse feedback, and distribution shift.
  • Efficiency trade-offs of ID representations, including embedding-table and codebook size, token-sequence length, training cost, and inference latency.

Deployment and Adaptation

  • Raw-ID embedding and semantic-code staleness as item content, interaction patterns, and catalogs evolve.
  • Incremental ID assignment and representation learning for new and changing items, including cold-start initialization.
  • Embedding refresh, semantic-code reassignment, and codebook or encoder updates, with trade-offs between freshness and representation stability.
  • Versioning, backward compatibility, and migration of ID features or tokens, preserving item resolution and historical interaction consistency.
  • Coordination of ID representation updates with model retraining, index maintenance, and serving infrastructure.
  • Continual and cross-domain adaptation of ID representations, drift monitoring, and experience from industrial deployments.

Submission Guidelines

We welcome technical studies, controlled comparisons, reproducibility work, negative results, deployment experiences, and position statements. Contributions may address raw, semantic, or hybrid ID representations within the workshop’s scope; both studies of individual designs and comparisons across ID families are welcome.

  • Review: the working plan is at least three non-conflicted Program Committee reviews per submission, with an unconflicted review lead handling disagreements and organizer submissions.
  • Proceedings: accepted papers can be published in the ACM Web Conference 2027 Companion Proceedings. Authors may opt out of proceedings publication.
Submission details to be announced

Paper lengths, formatting instructions, anonymity policy, submission platform, and final publication policies will be published when confirmed. Submissions are not open yet.

Important Dates

Paper submission deadline
January 4, 2027
Author notification
February 1, 2027
Camera-ready deadline
February 16, 2027
Workshop day
TBA · The Web Conference 2027
May 10–14, 2027 · Dublin, Ireland

Paper deadlines are Anywhere on Earth (UTC−12). The exact workshop date will be announced; session times will use the local time (GMT+1).