Entity disambiguation: sentence context hits 97.9% vs 63.5% for name matching
YvesMulkers · x · 2026-11-17
7wData's news knowledge graph kept mislinking companies: security firm Okta became the first name Oktay, machine-health firm Augury matched a 2021 video game, and FMR LLC (Fidelity's holding company) was confused with fMRI brain scans.
They benchmarked two approaches on 30,000 random name mentions from 2025-2026 news articles:
- Quick name matching (look up spelling, take the best-known entry): only 63.5% pointed to the correct Wikidata entry
- Deciding from surrounding sentences (choosing among Wikidata candidates or none): 97.9% correct
The fix: treat Wikidata as the public registry of real entities and link each mention to a unique entry; two articles pointing at the same entry are about the same company. The takeaway is "read the sentence, not the name," and the author offers a free kit with the prompts.
More from coding & agent
- Pageloop Launches a Docs Agent That Auto-Captures Screenshots and Updates Stale Docs — prasanna_says · 2026-11-17
- OutfitterBench tests LLM agents on 200-day rental business, exposing commitment failures — Jie Huang · 2026-11-17
- Craftax framework evolves code policies via feedback cascade, reaching final stage 50% of time — Víctor Gallego · 2026-11-17
- XState v6 Beta Launches: State Machines Rewritten for the AI Agent Era — mattpocockuk · 2026-11-17
- Matt Pocock shares a prompt for having AI agents evaluate new tech via draft PRs — mattpocockuk · 2026-11-17
- Motion launches Claude Code for Video Editing: raw footage plus one prompt, auto-edited — CurieuxExplorer · 2026-11-17