The workshop Smarter Extraction of ScholArly MEtadata using Knowledge Graphs, Language Models and Agents (SESAME) brings together researchers and practitioners interested in improving scholarly metadata through large language models, Agents, knowledge graphs, natural language processing, and linked-data technologies.

High-quality scholarly metadata is essential for search, discovery, research assessment, reproducibility, and the long-term operation of digital-library infrastructures. However, metadata about publications, authors, affiliations, datasets, software, models, citations, and other research objects is frequently incomplete, inconsistent, or distributed across heterogeneous sources.

SESAME focuses on methods that combine language models and structured knowledge to support reliable metadata extraction, normalization, enrichment, linking, validation, and evaluation. The workshop provides a platform for researchers, digital-library professionals, data curators, infrastructure providers, and policy experts to exchange methods, datasets, systems, evaluation practices, and lessons learned.

The workshop particularly welcomes work connecting large language models with linked data and knowledge graphs for tasks such as author disambiguation, affiliation normalization, citation-context understanding, persistent-identifier linking, provenance tracking, and transparent metadata curation.

Topics of Interest

Topics include, but are not limited to:

• Research Artifact Metadata Modeling and Granularity
  • Metadata of Scholarly Publications, Datasets, Software, and Models
  • Metadata Quality Assessment, Enrichment, and Curation
  • Research Artifacts Provenance across Digital Libraries
  • Cross-disciplinary Scholarly Metadata Interoperability
  • Persistent identifiers, including DOI, ORCID, ROR, and related identifier systems
• Large Language Modelsa and Agents for Natural Language Processing (NLP) for Scholarly Metadata
  • Research Artifacts Metadata Extraction using LLMs and Agents
  • Prompt Engineering, fine-tuning for Scholarly Information Extraction
  • Skill Development in Agents for Scholarly Information Extraction
  • Evaluation, Reliability, and Issues for LLM-generated Scholarly Metadata
  • Agentic approaches for Evaluating Scholarly Metadata
  • Comparative Studies of LLM-based vs Agentic vs Traditional Methods
  • LLMs and Agents for Scholarly Metadata Curation and Normalization
  • AI-driven Curation, Preservation at scale, and long-term accessibility
• Knowledge Graphs and Linked Data
  • Construction of Scholarly Knowledge Graphs from Heterogeneous Metadata
  • Linking and Aligning Entities across Repositories and Research Infrastructures
  • Applications of KGs for Discovery, Recommendations and Impact
• Digital Libraries and Research Infrastructures
  • Integration of Metadata workflows into Digital Library Systems
  • Benchmarks, Datasets, and Shared tasks for Metadata Extraction and Modeling
  • System Design for Metadata-intensive Digital Library Applications
• Societal, Ethical Impact and Future Policy Directions
  • Ethical Implications of AI-driven Metadata Generation and Curation
  • Metadata for Open Science, Reproducibility, and Research integrity
  • Societal Impacts of Metadata Granularity on Scholarly Evaluation and Equity
  • Policy Frameworks and Governance for Interoperable Metadata Infrastructures

Call for Papers

SESAME 2026 invites original research contributions addressing the above listed topics in three categories.

• Long Papers: 6–8 pages, excluding references
• Short or Position Papers: 2–4 pages, excluding references
• Demo, Dataset, or Benchmark Papers: 2–4 pages, excluding references

Important Dates (AoE)

All deadlines will use Anywhere on Earth (AoE) time.

• Paper submission: 15.08.2026
• Author notification: 01.09.2026
• Camera-ready submission: 10.09.2026
• Workshop date: To be announced

Submission

• Submission site: EasyChair
• Language: All submissions must be written in English.
• Format: All submissions must follow the CEUR Workshop Proceedings style.
• Review Policy: Double-blind peer review.
• Supplementary Material: Data, source code, and models are encouraged where appropriate.

Submissions should contain original work and should not be simultaneously under review elsewhere.