Status: Accepted
Date: 2026-09-26 (proposed), 2026-09-26 (accepted)
Related: ADR-A85 (scoped binding resolution), ADR-A87 (concept exclusion), ADR-A98, ADR-A100
Unit: applied-insurance-reference (AIR-0.1)
Sketches: peril-vocabulary.md, peril-structure-whitepaper.md
What a peril means in a contract is fixed by the contract’s drafter (a carrier, an MGA, a broker) or by its market, which may publish its own list. Consumers bring those lists, most of them flat or limited to a few levels. Lattice’s applied insurance domain still needs a reference peril vocabulary: a default where no other vocabulary is bound/available, to provide a target for crosswalks and the structure that contract checks need. The whitepaper shows that a SKOS hierarchy alone cannot carry that structure: kinds and parts differ, triggering is directed, some perils overlap without subsumption, cyber and flood wordings select by mechanism and agency, bundles are sets, and some perils are defined by thresholds.
Vocabulary resolves scoped bindings by one rule (ADR-A85): among applicable bindings, one whose scopes are a strict superset of another’s wins, and any remaining tie is a governance conflict. Bindings at agreement, organisation and market level therefore need a declared scope pattern, or they tie.
skos:editorialNote on each scheme, and binds each as voc:boundScheme on
its contract in insurance/common/. A reference edition is never presented as a market
standard. Any other edition, flat or structured, may be bound instead, and need not extend the
reference.Scopes nest. Bindings above the fallback name nested scopes, so that ADR-A85’s rule orders them without further logic:
| Binding | Names the scopes |
|---|---|
| market (a market-published edition, for example) | market, and a legal regime where the edition is regime-specific |
| drafting organisation | the market scopes, and the organisation |
| agreement | the organisation binding’s scopes, and the agreement |
A binding outside this pattern that ties with another is reported as a conflict, as Vocabulary already specifies.
Structure beyond SKOS is expressed as sub-properties and separate schemes, so a consumer reading only SKOS sees ordinary broader, related and semantic relations:
| Structure | Expression |
|---|---|
| kind and part | prl:broaderGeneric, prl:broaderPartitive ⊑ skos:broader, aligned to the thesaurus standard’s iso-thes:broaderGeneric and iso-thes:broaderPartitive. skos:broader is materialised |
| triggering | prl:canTrigger ⊑ skos:semanticRelation, irreflexive, neither symmetric nor transitive. SKOS has no directed associative relation, and skos:related is symmetric |
| overlap | prl:overlaps ⊑ skos:related, declared symmetric and irreflexive itself |
| integrity | neither prl:canTrigger nor prl:overlaps links two concepts in one skos:broaderTransitive chain, checked by shape |
| characteristics | one scheme per axis (mechanism, agency, onset, definition basis, accumulation class), linked from cause concepts for typical values and from occurrences for facts. The name is “characteristic”. “Facet” is not used |
| collections | skos:Collection sub-classes for bundles, standard sets and model groupings, expanded at bind time |
| thresholds and windows | links to Quantification: intensity measure, defining threshold as a qnt:RangeSet, customary event window |
insurance/peril/ is built in Phase 2 of the epic. Its layout is in the peril vocabulary
sketch §4, with paths per ADR-A98.