OSINT / IDENTITY

Match entities without merging namesakes

Two similar records do not necessarily describe the same object. Matching must retain sources, dates and decision reasons, including when no merge is possible.

Document verification desk with a ledger and company file.

AI-generated illustration

Choose the unit you need to identify

First state whether the record represents a person, legal unit, establishment, brand or research organisation. A relationship between objects is not identity. A normalised name helps find candidates; it is not a merge key.

Identifiers do not all identify the same object

Identifiers do not all identify the same object
IDIdentified objectCheck
SIRENLegal unitNine digits. Check the registry and store as text to preserve leading zeros. A trading name does not replace the identifier.
SIRETEstablishmentFourteen digits; the first nine are the SIREN. Two SIRETs with the same prefix may refer to one legal unit but to two distinct establishments.
ORCID iDPerson in researchRetain the full URI and leading zeros. The last character can be X. An ORCID identifies a person record; it does not automatically certify every profile statement.
ROR IDResearch organisationReview the candidate, location and relationships. An affiliation string may mix department, parent organisation and address; do not automatically pick the first similar name.
SIRENLegal unit
SIRET AEstablishment
SIRET BEstablishment
Link establishments to the legal unit — Keep establishment records, addresses and dates. Do not merge them into one location.

Five checks before merging

Same object and registry

Compare identifier system and value together, not an isolated character string. An establishment and its legal unit need linked records.

Source for each attribute

Retain the URL or document, observation date and known effective date for names, addresses and status. An aggregated profile does not replace the source document.

Compatible period

Different addresses may reflect a move; another name may be historical. Check periods before replacing an old value.

Visible contradictions

Different IDs in the same registry or incompatible places require review. Do not erase conflicts to tidy the record.

Reversible merge

Keep original row IDs, rule, reviewer, date and a mapping table. Later corrections must allow records to be separated.

A repeatable method

  1. Keep raw data

    Store original labels and identifiers separately from normalised values.

  2. Generate candidates

    Normalise spacing and case for searching. Preserve accents, alternative names and transliterations as variants; avoid dropping suffixes that distinguish objects.

  3. Compare independent records

    Check the ID in its registry and other attributes against appropriate sources. Two portals copying one registry are not independent confirmations.

  4. Classify the decision

    Use “documented same object”, “distinct objects”, “relationship to retain” or “undetermined”. Similarity is not certainty.

  5. Review the effect on the dataset

    Compare row counts, broken links and lost attributes. Review rejected matches too, to find missed duplicates.

Four common decisions

These are schematic cases, not accounts of real people or businesses.

Same name, different identifiers

Decision : Keep two candidates

Decisive check : Check registry, object and period; a name alone cannot decide.

One SIREN, two SIRETs

Decision : Link establishments to the legal unit

Decisive check : Keep establishment records, addresses and dates. Do not merge them into one location.

Same ID, old address

Decision : Retain the history

Decisive check : Check source and effective date; distinguish a real change from a typing error.

Department and parent organisation

Decision : Create a documented relationship

Decisive check : Check record granularity and institutional relationship. Affiliation is not identity.

Record to retain

For each decision retain: record IDs A and B; object type; registry and identifiers; raw labels; sources and dates; agreements; contradictions; reasoned decision; reviewer; revision date. A missing field means “not supplied”, not “false”.

Practical questions

Does a valid identifier format prove existence?

No. Format or checksum catches some typing errors. Check the official registry record and whether it identifies the intended object.

Is it possible to merge on an address or web domain?

Not automatically. Multiple organisations may share an address and one website may represent a group. Treat these fields as clues.

How should translated names be handled?

Retain published name, language, variants and a verified shared identifier. Similar transliterations do not prove two people are the same.

Do different IDs prove objects have no relationship?

No. They may identify distinct related objects or belong to different registries. Document the relationship instead of inventing equivalence.

Should all matched attributes be published?

No. Collection, internal analysis and publication are separate decisions. Keep attributes needed for the subject and adapt disclosure.

How is automatic matching evaluated?

Use reviewed cases with false matches and missed duplicates. Measure precision and recall within the test set’s stated scope; a similarity score is not a reliability rate.

DOCUMENTED REFERENCE

Wikidata: retain statements, qualifiers and references

A direct entity-to-value link does not carry all statement context. Wikibase RDF separates subject, statement node, qualifiers and references. This matters for dated roles, measurement methods and multiple values. Prefixes express graph roles, not trust levels.

wd:Q16wds:…?value?time?refp:ps:pq:prov:wasDerivedFromP1082 · P585
Wikidata: retain statements, qualifiers and references
ReferenceValue or formWhat it identifies
wd:EntityEntity identifier, such as Canada wd:Q16 in official examples. Translated labels do not replace identifiers for joins.
p:Entity → statementLinks an entity to a statement for a property. Several statements need not be duplicates.
ps:Statement → valueLinks a statement to its simple main value. Full value nodes can add quantity or time components.
pq:QualifierStatement qualifier. In the population example pq:P585 supplies point in time, not necessarily data-entry time.
prov:wasDerivedFromStatement → referenceLinks a statement to a reference node. Open the source and check actual support for the value.
pr:Reference propertyProperty inside a reference. Multiple references supporting one statement must not be counted as extra people or events.
wds:Statement identifierStatement-node identifier. Its internal format and meaning are not guaranteed; do not infer dates or business identity from the string.

Implication for the record

Choose entity, statement or reference as row unit before flattening. Multiple qualifiers and references can multiply rows for one statement. DISTINCT removes identical rows, not a wrong counting unit. Preserve query, execution date and identifiers.

Is it possible to keep just the entity and latest date?

Only with a justified selection rule. Dates may qualify measurements or historical periods; a maximum without correct statement links can combine unrelated values.

Primary documentation

Documentation consulted on .

DOCUMENTED REFERENCE

ORCID: inspect assertion origin and public-record limits

ORCID helps distinguish namesakes, but a record combines assertions from different origins. ORCID separates who adds an item from who connects it to the person. Holder-controlled visibility also limits what a public reader can observe.

ORCID: inspect assertion origin and public-record limits
ReferenceValue or formWhat it identifies
SourceWho added the itemSource is the person or system adding the item. An organisation source name indicates a deposit role, not intellectual authorship of all data.
Assertion originWho made the connectionAssertion origin made the connection to the holder. A researcher can initiate an import while a service technically adds the item.
AffiliationAssertion + sourceRead affiliation with source, organisation and available dates. Self-added and institution-added affiliations have different provenance paths.
PUBLICEveryonePublic items can be read in the public record and public API. Visibility is not independent proof of each claim.
LIMITEDTrusted partiesTrusted-party items are not part of the public record. Reading depends on permission; outsiders cannot infer their content.
PRIVATEOnly mePrivate items are not publicly visible. Documentation details holder, delegate and, where applicable, depositing-integration access.

Implication for the record

An empty or incomplete public record does not prove no employment, funding or publications. Preserve access date and do not reconstruct biography from visible fields alone. An authenticated ORCID workflow links an account, not automatic certification of every qualification.

Does an organisation source guarantee the whole record?

No. Inspect each item’s provenance and distinguish source from assertion origin. An institution-added affiliation need not validate articles separately added by the holder.

Primary documentation

Documentation consulted on .

DEEPER READING / PRACTICE

Document the evidence for an entity match

Two lines seem to describe the same organization, but their names or addresses are different. The approximation must be based on several attributes and preserve the initial data.

Updated on .

Priority checks

Compare stable identifiers, territory, period, name form and other available attributes. Examine cases of homonymity and names change before a merger.

Interpretation pitfalls

Aggressive normalization can make a significant difference disappear. A automatic approach must be able to be revised when the indicators are low or contradictory.

Working record to retain

The approximation journal retains original lines, rule, indications, decision and reserves. The ambiguous cases remain separate as long as an additional element does not justify their meeting.

PRACTICAL CASE / WORKED EXAMPLE

A common address leads to an incorrect fusion

The example below is pedagogical. It shows how to move from a document to a qualified proposal, preserving the context, the contradictory elements and the information that the research does not yet allow to establish.

Example context

Two organizations share an address and close names. A approach rule brings them together while their identifiers and their documented activities remain separate.

Suggested approach

Back to the original lines, compare several attributes and examine the available evidence. Separate case of proximity from case of identity. Documentation of the fusion corrections.

Conclusion and next step

The newspaper retains the revised decision and its reasons. The rule is adjusted so that the address serves as a complementary index rather than sufficient identity proof.

Questions for deeper practice

How to deal with uncertain approaches?

Keep them in a category to be examined, with the indicators for and against. This separation preserves the data and prevents unjustified fusion in the following analyses.

DETAILED WORKSHOP / APPLICATION

A rule on the cases of misleading similarity

  1. 01 / Prepare the necessary information

    Select attributes, identifiers and original data. Preparing known correspondences as well as homonyms or structures that share an address.

  2. 02 / Carry out the approach

    Test the rule, examine the indications and keep the decisions. Check the false approaches and missed cases. Document the transformations of the names or codes used for the comparison.

  3. 03 / Examine an alternative situation

    If a fusion depends on a single low index, keep the lines separate. An uncertain category allows a review without immediately imposing a common identity.

  4. 04 / Decide on the next step

    The rule is acceptable for its scope when major errors are examined. Manual and automatic decisions remain identifiable with their motives.

  5. 05 / Hand over the verification records

    Transmitting marriage journal and examples of control. A review can recover the original values and correct the analyses depending on an incorrect merger.

ANNOTATED WORKING DOCUMENT / REVIEW GRID

A book with indications for and against

The journal retains original lines, transformations, rules and decisions. It distinguishes confirmed correspondence, separation and uncertain case so that the approaches can be revised.

On a small screen, scroll the grid horizontally to read all three columns.

Example situations to assess
ObservationSupported interpretationNext step
Several stable attributes agree.The identity is supported according to the evidence and the documented context.Keep the reasons and check the limit cases of the rule.
A close name and a shared address agree alone.The approximation remains fragile in the face of homonymities and neighboring structures.Search for identificators or maintain an uncertain category.
The fusion produces attributes that are incompatible in the same period.The rule can gather separate objects.Back to the original and correct the dependent analyses.

Reviewing the working document

The final verification must be able to cancel a merger and recover the source values. Normalizations must be kept separately so that research aid cannot be converted into evidence of identity.

REPORTING / EXAMPLE WORDING

Presentation of the results: Documenting the indications of an approach

Point to clarify

A fusion decision must be able to be reviewed from the initial data.

Before handing over

Check the contradictory indications and the effects of the merger on the analyses.

Example of a qualified conclusion

The match relies on the stated clues. Keep ambiguous cases separate and retain attribute transformations together with their rules.