Building a knowledge graph for climate policy
summary
An open-source knowledge graph combines in-house policy expertise with scalable data infrastructure to label 27.5k documents (2M pages, 45M passages) and 350k annual users, using a Wikibase concept store with roughly 1,500 concepts authored by experts, a classifier factory that produces and promotes models (about 80 classifiers today, aiming for thousands), human validation plus a vibe checker for evaluation, and an inference pipeline feeding Vesper search orchestrated by Prefect that is reduced from 137 days to under 24 hours.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 0:32What is Climate Policy Radar
- 2:00Why a knowledge graph
- 7:40Defining concepts
- 11:23Classifiers
- 14:13Validation
- 17:26Vibe Check
- 17:58Scaling
- 18:49Text classification pipeline
- 19:20Training
- 21:15Inference and indexing
- 25:49Observability
- 29:40Results
- 30:28Questions
- 33:20Architecture
- 35:34Illustrations
related talks
44mBuilding Knowledge Graphs- 44mBuilding Knowledge Graphs
46mBuilding Knowledge Graphs
11mEnhancing Supply Chain Resilience and Transparency Through Spatial Knowledge Graphs
12mThe Green Potential of Knowledge Graphs
24mWhy You Need a Knowledge Graph And How to Build It
29mBuilding a knowledge graph with the Natural History Museum to work as a taxonomic equivalence engine
27mClimate and Compliance: The NGFS Climate Scenarios