.svg){kind=icon}
Learn
Record linkage is the statistical method for matching structured records that refer to the same thing; entity resolution is the broader capability that resolves entities across many structured and unstructured multi-source systems, preserving provenance for analyst review.
Last updated: October 18, 2026
Why it matters to your mission. Analysts lose hours deciding whether two records describe the same person, unit, or vessel, and a wrong call quietly corrupts every judgment built on top of it. Record linkage is the statistical method that scores whether two structured records match; entity resolution is the wider capability that resolves entities across many structured and unstructured systems at once and keeps a line back to every source. The outcome is one briefable, defensible picture instead of a pile of unmatched leads, assembled at machine speed with the reasoning an analyst can inspect.
Record linkage is the statistical method for matching structured records that refer to the same thing; entity resolution is the broader capability that resolves entities across many structured and unstructured multi-source systems, preserving provenance for analyst review. Record linkage is the technique; entity resolution is the operational capability that uses it, along with several others, to turn fragmented reporting into one trustworthy entity.
Put another way: all record linkage is a form of matching, but entity resolution is more than matching. Record linkage answers "do these two rows describe the same thing?" Entity resolution answers "across every system we hold, structured and unstructured, who is this entity, what connects to it, and how confident are we, with a trace back to every source?"
No. They are related but not interchangeable, and the distinction matters for anyone evaluating a capability.
Record linkage is a well-defined statistical discipline with a formal foundation. The U.S. Census Bureau's work on the subject describes how the field's decision rules were set out "to describe methods introduced by Newcombe" and formalized in "the seminal paper by Fellegi and Sunter (1969)," which remains the standard reference (see W. E. Winkler, Matching and Record Linkage, U.S. Census Bureau, 1993). The Fellegi-Sunter model compares fields between two records, weighs how much each agreement or disagreement contributes, and classifies the pair as a link, a non-link, or a possible link routed for clerical review. That is record linkage: a measurable way to match structured records.
Entity resolution is the broader operational capability. It uses record linkage as one of its techniques, but it also extracts entities from free text, generates candidates at scale, clusters many records into one entity, attaches a confidence score, routes ambiguous cases to a human, and persists the result with provenance so the match can be inspected and reversed. Record linkage is a method inside that capability; entity resolution is what you field to actually resolve entities across a real, fragmented data environment.
Hold the two apart this way: record linkage is a chapter in the textbook; entity resolution is the working system you deploy against the mission.
The difference shows up the moment the data stops being clean. Classic record linkage was built for a cooperative problem: two reasonably structured files, defined fields, and subjects who want to be matched. Entity resolution for intelligence and defense work inherits almost none of those comforts. The subject is often adversarial, the inputs include free text and signals and imagery-derived data with no shared schema, and the data spans classification levels that cannot be pooled into one file.
So record linkage handles the matching math on the records it can see, while entity resolution handles everything around the math: pulling entities out of prose, comparing across sources that were never designed to line up, preserving the originals rather than overwriting them, and keeping the whole chain auditable. Record linkage can tell you two rows probably match. Entity resolution tells you who the entity is across your entire holdings, what it connects to, and why you should believe the answer.
This matters because every downstream product inherits the result. The Department of Defense has made integrating data across hundreds of separate systems a program-level priority through its Combined Joint All-Domain Command and Control (CJADC2) effort, and independent oversight has documented how hard that is in practice: analysts still "collect, assess, and package the data by manually entering and transferring them into yet another system," and the goal is a data integration layer that "effectively connects existing federated data across multiple systems and mission areas" (U.S. Government Accountability Office, GAO-25-106454, Defense Command and Control, 2025). Record linkage alone does not solve that; entity resolution, run across all those federated sources with provenance intact, is the step that does.
The table below separates the two so each can be evaluated on its own terms. Record linkage is a technique; entity resolution is the capability that operationalizes it across a multi-source environment.
| Dimension | Record linkage | Entity resolution |
|---|---|---|
| What it is | A statistical method for matching records | The operational capability that resolves entities across systems |
| Scope | Pairs or files of structured records | Many sources at once, structured and unstructured |
| Input data | Defined fields in structured tables | Structured records, free text, signals, and imagery-derived data |
| Core question | Do these two records match? | Who is this entity across everything we hold, and what connects to it? |
| Foundational basis | The Fellegi-Sunter model (field agreement weights, link / non-link / possible link) | Record linkage plus extraction, candidate generation, clustering, scoring, and adjudication |
| Handling of free text | Not addressed; assumes fields | Extracts entity mentions from prose before matching |
| Output | A match decision, often a merged record | A linked entity with confidence and provenance, originals preserved |
| Human role | Clerical review of possible links | Analyst adjudication of ambiguous matches, confidence surfaced |
| Auditability | Match weights are inspectable | Every resolved entity traces back to each contributing source |
Neither column is wrong. Record linkage is the rigorous matching core; entity resolution is what you need when the data is heterogeneous, partly unstructured, classification-constrained, and the subject is working not to be matched.
Use record linkage when the problem is two or more structured files that need to be matched or deduplicated against defined fields, and you need a defensible, measurable match decision. It is the right tool for a bounded, structured matching task, and its statistics are mature and well understood.
Reach for entity resolution when the real problem is larger: data arrives from many systems that were never designed to line up, a meaningful share of it is free text or signals or imagery-derived, the same entity hides behind aliases and transliterations, classification limits what can be compared, and every resolved entity has to survive a briefing. In that environment matching is necessary but not sufficient. You need the surrounding capability, and you need it to preserve the sources rather than collapse them into one golden record.
In most defense and intelligence settings the honest answer is that record linkage is one engine inside the entity resolution you have to field. Evaluating record linkage in isolation understates the problem, because the hard part is rarely the matching math; it is the heterogeneity, the provenance, and the adjudication around it.
Torch.AI builds government-owned reasoning infrastructure the customer can own and govern. Within it, NEXUS makes data meaningful: it extracts entities from unstructured text and gives multi-source records a common semantic representation, which is the step record linkage alone never handled. HALO then performs entity resolution and graph-based fusion, resolving records that describe the same real-world entity across fragmented, multi-source systems, attaching confidence, and preserving a trace back to every source so the match can be inspected and reversed. The matching math of record linkage lives inside that capability; it is not the whole of it.
This reasoning layer complements existing systems of record rather than replacing them, so authoritative data stays intact and interoperable instead of locked inside a proprietary semantic model. Whether a specific deployment is government-owned (GOTS) or commercial (COTS) depends on the system the customer installs and purchases; the point is that the government-owned model is available to evaluate when ownership and control matter to the mission. You can see how this is packaged as a capability on the Torch.AI software page. For the full capability, see entity resolution for intelligence analysis.
If you are evaluating whether a tool gives you record linkage or genuine entity resolution, require the following:
Scoping a capability? Bring these seven requirements to a call and we will walk each one against your environment and run it on a sample of your multi-source data, with provenance traced end to end: request a technical walkthrough.
No. Record linkage is the statistical method for matching structured records, formalized in the Census Bureau's Fellegi-Sunter model; entity resolution is the broader operational capability that uses record linkage alongside extraction, clustering, confidence scoring, provenance, and analyst adjudication to resolve entities across many structured and unstructured sources.
Record linkage handles the matching math on the structured records it can see; entity resolution handles everything around it: extracting entities from free text, comparing across sources with no shared schema, respecting classification, preserving the originals, and keeping the whole chain auditable. Record linkage says two rows probably match; entity resolution says who the entity is across all your holdings and why you should believe it.
Use record linkage for a bounded task of matching or deduplicating structured files against defined fields. Use entity resolution when data spans many systems, a meaningful share is unstructured, aliases and classification constraints are in play, and every resolved entity has to survive a briefing. In most defense settings record linkage is one engine inside the entity resolution you actually field.
It is the foundational framework for record linkage, set out in the 1969 paper by Fellegi and Sunter and documented by the U.S. Census Bureau. It compares fields between two records, weighs each agreement and disagreement, and classifies the pair as a link, a non-link, or a possible link routed to human review. Modern entity resolution builds on it rather than replacing it.
No. Entity resolution incorporates record linkage as its matching core and adds the capabilities a real multi-source environment demands. The matching statistics carry over; what changes is everything around them.
Describe the problem.
Define the constraints. We will structure the session.
No demo.
No slide deck.
Real work.
Describe the problem.
Define the constraints. We will structure the session.
No demo.
No slide deck.
Real work.
