Cognate vs Loanword Classifier

Independently documented historical inheritance vs borrowing/contact evidence for cognate/loanword relation; surface spelling is not proof. Independently declared positive requirements, overlapping/unknown classes retained. Caller establishes language, dialect, method and reference context; no automatic identity or universal language-wide claim.

Description

Cognates and loans involve different historical relationships between forms.

Cognates and loans involve different historical relationships between forms.^[1]

Assessing them requires positive etymological and contact evidence under declared criteria. The tool compares your observed and known flags with your declared positive requirements, retaining unresolved candidates.

When to use Cognate vs Loanword Classifier

Use this tool to calculate or review the supplied observations under the stated rules. Specify the language variety, analysis, source edition and population or inventory definition.

How to use Cognate vs Loanword Classifier and understand its results

Specify the language variety, analysis, source edition and population or inventory definition.
Enter callerEtymologyRelationClassIds, callerEtymologyRelationRequirements, observedEtymologyRelationEvidence, knownEtymologyRelationFlags, using the field meanings and rules below, then run the tool.
Read compatibleCallerClasses, fullyWitnessedCallerClasses, using the output meanings and calculation rules below. Distinguish compatible candidates from fully witnessed classes; compatibility can retain unknown required evidence.
Correct any reported input issue and run again. Before using a successful result, check the stated units, signs, reference scope and observation coverage.

Input field	Meaning	Value type	Requirement
callerEtymologyRelationClassIds	Caller etymology relation class ids	list of string	Required
callerEtymologyRelationRequirements	Caller etymology relation requirements	numeric matrix	Required
observedEtymologyRelationEvidence	Observed etymology relation evidence	list of number	Required
knownEtymologyRelationFlags	Known etymology relation flags; binary 1 means the matched observation is known, 0 means unknown	list of number	Required

Use unique nonblank class IDs and a nonempty binary requirement matrix aligned by class and evidence column. Every class must require at least one positive feature. A one requires presence; zero ignores the feature and does not require absence. Supply binary observed evidence and known flags. A known required zero excludes a class; an unknown required feature retains compatibility but prevents full witnessing. Return all compatible classes and the subset with every required feature known and present. Classes may overlap.

compatible_c=∀j(requirement_cj=0 or known_j=0 or evidence_j=1); witnessed_c=∀j(requirement_cj=0 or (known_j=1 and evidence_j=1))

Use the matching field units and the exact boundary, ordering and reference rules above.

Output field	Meaning	Value type
compatibleCallerClasses	Compatible caller classes	list of string
fullyWitnessedCallerClasses	Fully witnessed caller classes	list of string

json

{
  "input": {
    "callerEtymologyRelationClassIds": [
      "a",
      "b"
    ],
    "callerEtymologyRelationRequirements": [
      [
        1,
        0
      ],
      [
        0,
        1
      ]
    ],
    "observedEtymologyRelationEvidence": [
      1,
      0
    ],
    "knownEtymologyRelationFlags": [
      1,
      0
    ]
  },
  "output": {
    "compatibleCallerClasses": [
      "a",
      "b"
    ],
    "fullyWitnessedCallerClasses": [
      "a"
    ]
  }
}

Limitations and assumptions

The language variety, analytical definitions and reference scope are supplied by the caller. No corpus, speech or complete/current linguistic registry is analyzed automatically. Compatible classes with unknown evidence are not positive identifications. Fully witnessed classes satisfy only the declared positive requirements.

Missing required fields, unknown fields, wrong types and violated rules return input issues. Numeric values and outputs must be finite. Lists and matrices must be dense, without missing cells. Numeric lists allow at most 10000 entries; ordinary string lists allow at most 128 strings of 128 UTF-16 units each; numeric matrices allow at most 128 rows and 128 columns, subject to narrower requirements above.

Alternative or Complementary methods

For a related step, use False-Friend Pair Examples Lookup: False friends are similar-looking expressions with potentially misleading meanings across languages. For a complementary comparison, use Calque Example Pair Lookup: A calque translates the components or structure of an expression rather than borrowing its entire form.

References

Cognate vs Loanword Classifier — tool.io

Similar or alternative tools

False-Friend Pair Examples Lookup
False friends are similar-looking expressions with potentially misleading meanings across languages.
Calque Example Pair Lookup
A calque translates the components or structure of an expression rather than borrowing its entire form.