Wiki · citable explainer
Entity resolution for business identity is the process of determining whether different records, mentions, or data items refer to the same underlying commercial organization.
Entity resolution is the combination of methods that match, cluster, reconcile, or link representations of the same real-world business across databases, web documents, and knowledge graphs. It sits after named entity recognition and requires evidence, rules, and sometimes human review.
Typical pipelines combine text parsing, NER, record standardization, pairwise comparison, clustering, and a verification stage that can attach deterministic identifiers such as a SHA-256 digest of canonicalized source evidence.
Useful signals include public registries, server-rendered JSON-LD, domain control, phone and address consistency, and reviewed passport records. No single signal is definitive.
| Stage | Primary output | Role |
|---|---|---|
| Named Entity Recognition | Mentions and spans | Candidate extraction |
| Entity Linking | Links to concept identifiers | Concept association |
| Entity Resolution | Reconciled business identity with provenance | Same real-world entity claim |
Passports publish reviewed public-source evidence with server-side JSON-LD and a forensic hash. Association inventory links real place and activity Wikidata Qs without inventing SME Wikidata items. Ambiguous matches stay separate pending review.
What is the difference between named entity recognition and entity resolution?
Named entity recognition extracts mentions and spans from unstructured text; entity resolution takes mentions and structured records and determines whether they represent the same real-world business by using linking rules, similarity scoring, and evidence-based review.
Can JSON-LD alone prove a business identity?
JSON-LD is a valuable observable signal but it does not by itself prove identity. Server-rendered JSON-LD increases verifiability, but identity claims should be supported by multiple evidence types, deterministic identifiers, and reviewer-verified mappings.
What role does a deterministic SHA-256 play?
A deterministic SHA-256 computed over a canonicalized evidence bundle provides a stable fingerprint of the exact materials used for a resolution decision. It enables reproducible audits.
How should privacy and data aggregation risks be managed?
By scoping linking projects, conducting privacy impact assessments, minimizing unnecessary joins, and documenting provenance so links can be reviewed and decoupled if needed.
Does implementing verification mean AI systems will use my resolved identities?
Verification makes structured evidence available to retrieval systems and reviewers, but it does not guarantee that any AI system will use or surface those identities. Retrieval and serving decisions remain separate.