KEIR @ ECIR 2025: The Second Workshop on Knowledge-Enhanced Information Retrieval

2025-01-20

AI Summary

Pretrained language models (e.g., BERT, GPT-4) suffer from static internal knowledge, limiting their ability to leverage real-time or domain-specific external knowledge—resulting in coarse-grained semantic understanding, biased relevance estimation, and poor domain adaptability. To address this, we propose Knowledge-Enhanced Retrieval (KER), the first systematic IR framework enabling dynamic external knowledge injection. KER integrates knowledge graph embedding, retrieval-augmented generation (RAG), entity linking, domain-adaptive fine-tuning, and multi-source knowledge alignment—moving beyond conventional fine-tuning paradigms. The project delivers an open-source toolkit, a benchmark dataset, and a comprehensive technical roadmap. Empirical evaluations demonstrate substantial improvements in cross-domain retrieval accuracy and interpretability. KER establishes both a methodological foundation and practical infrastructure for next-generation, knowledge-driven information retrieval systems.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Knowledge Representation and Reasoning: Knowledge Acquisition
• Natural Language Processing: Information Extraction

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Pretrained language models (PLMs) like BERT and GPT-4 have become the foundation for modern information retrieval (IR) systems. However, existing PLM-based IR models primarily rely on the knowledge learned during training for prediction, limiting their ability to access and incorporate external, up-to-date, or domain-specific information. Therefore, current information retrieval systems struggle with semantic nuances, context relevance, and domain-specific issues. To address these challenges, we propose the second Knowledge-Enhanced Information Retrieval workshop (KEIR @ ECIR 2025) as a platform to discuss innovative approaches that integrate external knowledge, aiming to enhance the effectiveness of information retrieval in a rapidly evolving technological landscape. The goal of this workshop is to bring together researchers from academia and industry to discuss various aspects of knowledge-enhanced information retrieval.

Problem

• Information Retrieval
• Semantic Understanding
• Domain-specific Knowledge

Innovation

• External Knowledge Integration
• Domain-specific Knowledge
• Semantic Understanding Enhancement