Large Language Models (LLMs) store extensive knowledge within their parameters, easily accessible through natural language interaction. However, they struggle when probed for long-tail knowledge (information rarely encountered during training). Conversely, Knowledge Graphs (KGs) excel at structuring specialized information but are often incomplete. Leveraging the parametric knowledge of LLMs and the authoritative knowledge stored in KGs could advance long-tail knowledge extraction and enhance its exploitation.

The first edition of “X-TAIL, eXtraction and eXploitation of long-TAIL knowledge” aims to attract researchers and practitioners operating at the intersection of KGs and Generative AI. X-TAIL offers an opportunity to engage in interdisciplinary discussions focusing on non-standard sources or working on methods and tools designed to aid in such scenarios.

Submissions will be published in CEUR-WS proceedings.

The main topics of interest are:

• Knowledge Extraction from non-standard unstructured sources:
  • Relation extraction
  • Entity Linking for long-tail entities
  • KG completion techniques for long-tail entities
• Knowledge Probing:
  • Investigations on biases hindering memorisation and generalisation capabilities of LLMs
  • Knowledge retention biases mitigation through KGs
• Development and combined use of LLMs and KGs in domain-specific data, for instance, but not limited to:
  • Cultural Heritage preservation
  • Digital humanities, sustainability, and industry-related applications
• Multi-modal Retrieval-augmented Generation (RAG) techniques:
  • Advanced question answering, search engines, content creation and summarisation, conversational agents and chatbots, information retrieval and knowledge engines with RAG on structured and unstructured sources
  • Natural language queries over KGs through RAG
  • Text generation systems based on KGs and/or LLMs through RAG

Submission format and guidelines:

• Papers must be submitted in PDF format according to the CEUR-WS template published in the CEUR-WS guidelines.
• Long papers should be between 10 and 15 pages, including references.
• Short papers should be between 5 and 9 pages, including references.
• Workshop papers must be self-contained and in English.
• At least one author of each accepted workshop paper has to register for the conference.
• Workshop attendance is only granted to registered conference participants.

• Abstract Registration Deadline: September 8th, 2024 September 13th, 2024
• Workshop Papers Submission Deadline: September 17th, 2024 September 24th, 2024 September 27th, 2024
• Workshop Papers Notification: October 15th, 2024
• Early Bird Registration: October 17th, 2024
• Workshop Papers Camera Ready: November 20th, 2024
• Workshop Day: November 26th, 2024
• Conference Days: November 26-28th, 2024

All deadlines are to be considered 23:59 AoE.

• Aldo Gangemi, University of Bologna, Italy
• Andrea Schimmenti, University of Bologna, Italy
• Andrea Zugarini, expert.ai, Italy
• Angelo Salatino, Open University, UK
• Antonello Meloni, University of Cagliari, Italy
• Benno Kruit, Vrije Universiteit Amsterdam
• Bohui Zhang, King's College London, UK
• Célian Ringwald, Inria Université Côte d'Azur, I3S, CNRS
• Chiara di Bonaventura, King's College London, UK
• Delfina Sol Martinez Pandiani, Centrum Wiskunde & Informatica, Netherlands
• Diego Reforgiato, University of Cagliari, Italy
• Gianmarco Pappacoda, University of Bologna, Italy
• Harald Sack, FIZ Karlsruhe, Germany
• Jan-Christoph Kalo, University of Amsterdam, Netherlands
• Mahsa Vafaie, FIZ Karlsruhe, Germany
• Nicolas Lazzari, University of Pisa/University of Bologna, Italy
• Rocco Tripodi, University of Venice, Italy
• Stefano De Giorgis, CNR Catania/University of Bologna, Italy
• Tabea Tietz, FIZ Karlsruhe, Germany

The workshop is co-located with EKAW-2024 and will be held on November 26th in Amsterdam. For more information about the exact location of the workshop, please visit this link.

For any inquiries, please contact us at X-TAIL workshop email address: xtailworkshop@gmail.com.