Constructing knowledge graphs without large language models (LLMs) is both feasible and cost-effective. Using the Relik framework from Sapienza University, which includes coreference resolution, entity linking, and relationship extraction models, you can efficiently build high-performance knowledge graphs. The setup involves using Python environments, Neo4j for storing graph data, and integrating smaller, task-specific models to reduce costs. The process results in structured knowledge graphs suitable for complex data analysis and retrieval tasks.
Table of contents
Combine coreference resolution, entity linking, and relationship extraction to build a knowledge graph without an LLM for your RAG applicationsEnvironment SetupNeo4j Aura - Fully Managed Cloud SolutionDatasetCoreference ResolutionEntity Linking and Relationship ExtractionQuestion AnsweringSummarySort: