Building a knowledge graph with the Natural History Museum to work as a taxonomic equivalence engine
summary
A knowledge-graph taxonomic equivalence engine maps US and UK taxonomies in TypeDB, loading ~2.5M taxa and ~5.5M relationships into a dockerized image that spins up in ~30s and takes ~20 minutes to load. Schema models synonyms as separate nodes with a concept description relation; rule-based inference (red/green match + transitive traversal) produces a closest node match when no direct species equivalent exists. Deployed via SageMaker, the pipeline cuts manual equivalencing from a day to ~10 minutes.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:05Natural History Museum need
- 6:26Naming challenges in data
- 8:57Taxonomic equivalence engine
- 9:19Knowledge graph benefits
- 10:22UK versus US taxonomy
- 11:46Modeling taxonomy in database
- 13:55Synonyms in UK taxonomy
- 16:46Direct taxonomy matching
- 18:33Inferring knowledge with rules
- 21:03Closest node match
- 22:52Loading and scaling graph
- 23:56Dockerizing solution
- 24:39Deploying MVP on SageMaker
- 25:42Business impact gains
- 26:27Lessons learned
- 27:32AI applications of knowledge graphs
related talks
44mBuilding Knowledge Graphs- 44mBuilding Knowledge Graphs
38mBuilding a knowledge graph for climate policy
22mThe Nexus of Innovation: How Knowledge Graphs Fuel AI Scientific Breakthroughs
42mBuilding a knowledge graph for climate policy
46mBuilding Knowledge Graphs
33mSmarter Industrial Fault Diagnosis with Knowledge Graphs and Large Language Models
41mData Science Education. Building Knowledge Graphs