Classification with a point of view

Begin with concepts, not an inventory of physical instruments#

OMARO 0.1.0 brings together 643 Hornbostel–Sachs classifications and 2,724 resolved instrument concepts across thirteen languages, with five unresolved stubs retained explicitly. These are vocabulary and assertion counts for that release. They are not counts of distinct surviving instruments. A label such as “double bass” denotes a concept that can refer to many physical objects with different histories and uses.

The source layer preserves MIMO-derived classifications and mappings, including the original exact-match statements where supplied. Retaining a source statement records that the source made it; it does not automatically identify a vocabulary concept with every physical object or endorse an unconditional equivalence. This distinction allows a reader to inspect the source without collapsing its interpretation into the ontology’s own strongest claim.

The same name can support more than one classification#

The release uses “Double bass” to illustrate multiple source classifications, including 321.322 and extensions concerning bare-finger and bowed use. The meaningful distinction is the sounding realization or specified playing technique. A classification appropriate to bowing is not thereby a claim that all performances of all double basses use a bow. The assertion needs an explicit perspective and applicability scope.

“Alto trombone” illustrates another issue: source mappings include the slide family and a short-air-column valve category. A name alone may leave construction underspecified. Preserve the source occurrence, then model the construction or context needed for a reviewed assertion. The goal is to make the disagreement inspectable, not to erase it by choosing whichever code appears first.

A source assertion is different from a reviewed projection#

The documented direct-projection policy requires two independent authorized reviews and no veto for an eligible context-independent claim. In release 0.1.0, classification_targets is intentionally empty. That is a meaningful publication state: the source and contextual evidence are available, while no relation has been silently promoted into the stronger direct view. An empty reviewed view does not mean the classification data failed to load.

Inspect the supplied SQLite vocabulary#

-- Look up an exact preferred label in German.
SELECT concept_uri, label
FROM preferred_labels
WHERE language = 'de' AND label = 'Signalhorn';

-- Inspect the reviewed direct projection.
-- An empty result is expected in release 0.1.0.
SELECT * FROM classification_targets;

These queries target the OMARO database, not the Pipe Organ Dataset. Begin with an exact language and label; keep the returned concept identifier when following assertions. In multilingual work, record the label used for discovery separately from the identifier used for analysis. A translated display label can change without changing the concept being examined.

Connect vocabulary and evidence conservatively#

OMARO is related to the MODAVIS Ontology Network without asserting a blanket ontology import or automatic compliance equivalence. Use it to ask richer questions: which classification is stated, by whom, about what kind of sounding realization, and under which conditions? In a dataset or interface, make those contextual distinctions available before presenting a single category as settled.