Merging the relational, the graph and the semantic
summary
Open-source Grakn knowledge graph implements a hyper-relational model with entities, roles and hyperedges and enforces an application metaschema checked at commit time so data and schema validity are guaranteed. It provides a human-readable declarative language called grackle with built-in rule reasoning (resolution) to produce virtual facts without persisting them. Storage delegates to Cassandra backend with gRPC streaming and a middle layer maps high-level graph constructs to key/value shards, enabling horizontal scaling; the system is reported as working in production.
AI-written summary · about
chapters (21)
- 0:00Introduction to CosmoGraph
- 1:11Inspirations behind the system
- 5:14Cognitive systems overview
- 7:28Key challenges in data systems
- 10:50Problems with unstructured data
- 13:03Database model requirements
- 14:33Approaches to data modeling
- 18:39Existing technologies overview
- 20:08Introducing Dragon database
- 21:39Ensuring data consistency
- 23:08Metaschema and schema definition
- 26:04Hybrid graph model
- 27:52Query language design
- 29:21Defining schema and data
- 31:11Adding data to the database
- 32:17Inheritance in schema
- 33:05Reasoning with rules
- 35:09Scalability architecture
- 36:39Communication and APIs
- 38:08Summary and conclusions
- 39:35Q&A session
related talks
01h 06mDesign and Implementation of the RelationalAI Knowledge Graph Management System
44mBuilding Knowledge Graphs- 44mBuilding Knowledge Graphs
25mAdaptive GraphRAG: A Framework for Knowledge Graph Quality, Consistency, Evolution
46mBuilding Knowledge Graphs
24mWhy You Need a Knowledge Graph And How to Build It
39mEnhancing RAG with Neo4j Knowledge Graph
41mGraphGen: Conducting Graph Analytics over Relational Databases