Datashare + Neo4j: Unstructured Docs to Knowledge Graphs for Investigative Journalism
summary
Datashare extracts text and named entities from terabytes of leak files (millions of docs from Panama/Pandora Papers) and a Neo4j plugin creates a graph from Datashare entities in one click so documents, people and companies become nodes; the exported graph is connected through documents not direct relationships and is often noisy, unresolved duplicates requiring ETL, local entity resolution and manual validation. Constraints: sensitive data stays on-prem, limited budget, and full-scale LLMs are impractical.
AI-written summary · about
chapters (5)
- 0:00Introduction and background
- 4:29DataShare and Neo4j integration
- 7:02Neo4j plugin for knowledge graphs
- 9:53Current limitations and goals
- 12:50Future roadmap and constraints
related talks
52mUsing Open Source Tech to Unearth the Stories of the Panama Papers
40mThe Paradise Papers: Behind the Scenes
17mMapping Argentina’s Scientific Ecosystem: Building Collaboration Networks with Neo4j
28mScaling Security Anomaly Detection in Enterprise Knowledge Base with Dask
32mCode More, Draw Less, and Debug Just a Little!
44mData Journalism 1
39mEmphasising Relationships in the BBC's Data Using Technologies of the Semantic Web
40mArchitectures for Exploring Your Personal Data