Introduction & Motivation
Modern AI systems process enormous volumes of multilingual text describing events across news media, social platforms, legal documents, and scientific literature. While transformer-based language models have vastly improved information retrieval, they often lack explicit mechanisms for representing event semantics, reasoning over complex event relationships, and explaining attribution decisions.
Event-Centric Attribution using Knowledge Graphs(KGs) bridges this gap by leveraging structured event networks to trace, verify, and explain LLM-generated claims, anchoring them directly back to real-world event structures to mitigate dynamic factual errors and curb hallucinations.Instead of an LLM producing an ungrounded or static response, an event-centric attribution framework links the output directly to a contextualized event node and its associated arguments within a Knowledge Graph.
[LLM Output] – Attributed to –> [[Event] Node in KG] [Event] |–> (Actor) Who |–> (Time) When |–> (Loc) Where
Key Research Advantages:
- Resolves Temporal Ambiguity: Grounds changing facts within precise, structured temporal scopes.
- Enhances Explainability (XAI): Offers human-auditable inference paths for tracking complex narratives.
- Supports Multi-Hop & Causal Reasoning: Facilitates logical processing across sequential and causal chains.(e.g., Event A triggered Event B, leading to Event C).
The workshop aims to produce collaborative research initiatives, and contribute towards encouraging research to produce event ontologies, semantic reasoning, multilingual understanding, structured event and provenance modeling, transparent attribution and evidence paths.
Scope & Topics of Interest
The workshop advances event-centric knowledge representation across core technical architectures, structural traditions (such as Indian traditional knowledge representations, Conceptual Graphs, UNL, and WordNets), and real-world deployment cases.
- Event-centric KG Construction
- Cross-document Event Coreference
- Event Temporal & Causal Reasoning
- Explainable Event Reasoning
- Graph Neural Networks for Event Analysis
- Multilingual Event Understanding
- Responsibility & Evidence Attribution
- Trustworthy Event Analytics
We invite researchers, academicians, and industry practitioners to submit original, unpublished research papers, case studies to this Workshop on Event-Centric Knowledge Graphs for Explainable AI.
We welcome submissions addressing core methodologies, open datasets, and neuro-symbolic architectures that utilize structural event networks to trace and ground large language model decisions. Accepted papers will be presented during our interactive research tracks and included in the workshop proceedings. Please ensure all submissions conform to the standard academic formatting guidelines as given in ICON 2026 and are submitted through softconf submission portal.
Tentative Schedule
The program seamlessly blends foundational technical keynotes with interactive peer-reviewed paper tracks and practical applications.In this we plan to conduct a Hands-on Tutorial session in which participants will learn to develop or build systems which can do Event-Relation extraction and provide explainable event attribution.
Organizing Committee
The organizing group brings extensive computational expertise from the Computational Linguistics Research Group (CLRG) at AU-KBC Research Centre.
More details about their work and publications can be seen on the below given website. AU-KBC CLRG Group Some of the relevant publications to this workshop are:
- RK Rao Pattabhi and Lalitha Devi Sobha. 2023. Chemxtract - A System for Extraction of Chemical Events from Patent documents. In Proceedings of Recent Trends in Natural Language Processing (RANLP), 2023.
- RK Rao Pattabhi and Lalitha Devi Sobha. 2023. Event Extraction from Social Media Text in Malayalam using Neural Conditional Random Fields. In Proceedings of FOSS Approaches towards Computational Intelligence and Language Technology (FOSS-CILT 23).
- RK Rao Pattabhi and Lalitha Devi Sobha. 2018. "EventXtract-IL": Event extraction from newswires and social media text in indian languages - 2nd edition - fire 2018 - an overview. In Proceedings of the Forum for Information Retrieval and Evaluation 2018, Gandhinagar, India.
- Lalitha Devi Sobha, RK Rao Pattabhi, and R Vijay Sundar Ram. 2018. "nigazaayvi" Event-Entity Profiling using Deep Syntactic and Semantic Analysis of Tamil Documents. In Proceedings of 17th Tamil Internet Conference, Coimbatore, India. Demo Link
- Gopalan Sindhuja and Lalitha Devi Sobha. 2017. Cause and Effect Extraction from Biomedical Corpus. In Proceedings of 18th International Conference on Computational Linguistics and Intelligent Text Processing (CICLing 2017), Budapest, Hungary.