Knowledge Graphs for RAG and Textual Document Analysis (Doc2KG 2026)

Goal of the workshop

The goal of this workshop is to demonstrate how integrating Knowledge Graphs with RAG overcomes the limitations of traditional systems and to explore how to build an architecture that enables complex reasoning and unlocks deeper, more accurate insights from documents. This workshop will aim to answer complex questions and generate breakthrough ideas.

Description

In today’s digital landscape, we are witnessing an unprecedented explosion in textual data generation. From social media posts and news articles to legal documents, academic papers, and business communications, the volume of text is growing exponentially. Traditional methods of analyzing these vast document collections frequently fall short in terms of scalability, accuracy, and efficiency, creating an urgent need for more sophisticated approaches.

The emergence of Retrieval-Augmented Generation (RAG) marked a significant advancement by grounding Large Language Models LLM in relevant contextual information. However, conventional RAG systems that rely primarily on vector similarity search reveal critical limitations when handling complex, real-world documents. They struggle with multi-hop reasoning connecting disparate facts across multiple documents and fail to capture the rich semantic relationships between entities such as people, organizations, and projects. This results in incomplete answers, factual inconsistencies, and an inability to perform genuine analytical tasks, leaving substantial untapped potential within corporate knowledge bases. This workshop introduces a paradigm shift: the integration of Knowledge Graphs (KGs) into the RAG pipeline. We will explore how representing document content as a structured, interconnected graph of entities and relationships can dramatically enhance the accuracy, depth, and reasoning capabilities of Large Language Models. The representation of data as graph structures has been empirically proven to significantly improve RAG performance, enabling more sophisticated document analysis. Doc2KG Workshop aims to bring together experts from industry, research, and academia to exchange ideas and discuss ongoing innovations in natural language processing and Generative AI for textual document analysis. Participants will gain comprehensive understanding of a cutting-edge architecture where documents are not merely embedded but transformed into dynamic graphs of interconnected entities. We will learn how this structured knowledge base enables precise, relationship-driven retrieval, allowing LLMs to traverse connections and deliver answers with enhanced accuracy, deeper context, and robust reasoning capabilities previously beyond reach.

The Doc2KG workshop aims to bring together an area for experts from industry, science, and academia to exchange ideas and discuss ongoing research in natural language processing and GenAI for textual document analysis.

The Doc2KG workshop encourages the participation of persons with disabilities, and underrepresented minorities in the STEM and competitive STEM workforce. Also, it encourages original application with a significant impact on the well-being of individuals in society. Finally, it greatly impacts increasing partnerships between academia and industry.

Scope of the workshop

Researchers and practitioners all over the world, from both academia and industry, working in the areas of document and textual analysis. Topics of interest include, but are not limited to, the following:

● Text to KG: Enhancing KG construction and completion with GenAI
● From KG to Text
● From Speech to text to KG
● Knowledge Graph Construction & Storage
● Document Ingestion & Pre-processing for KG construction
● Innovative pipeline for Knowledge Extraction
● Hybrid Retrieval & Querying of KGs
● Reducing Factual Hallucinations using RAG and KGs
● Prompting Engineering using KGs
● KG augmentation from document
● Triples representation for KG
● Specific domain KG querying
● Benchmark datasets relevant for tasks combining KGs and GenAI
● Real-world applications on scholarly data, biomedical domain, etc.
● Industry application and real-world scenarios application
● KG for legal text