MiKaDo

Purpose

Heritage knowledge is produced by people who look, measure and interpret while they record. The archaeologist drawing a section, the conservator marking an alteration and the conservation scientist choosing a sampling point bind observation, description and analytical intent in one gesture. Geometry of the object is routinely preserved; the conditions under which an observation was formulated — who acted, with which instrument, toward which question, on which region of the surface — stay queryable when the vocabulary keeps that passage.

MiKaDo is a model of mediation. Knowledge about heritage, in field and laboratory campaigns, is produced through situated, instrumented acts of observation. The role of the vocabulary is to articulate the passage from a material referent to shareable knowledge through those acts. Four corollaries follow. Knowledge is produced: every knowledge node is the outcome of an activity. It is situated: anchored to a place and to a time. It depends on an act: the walk from raw media to a statement always passes through an observed feature. It is plural: the same material referent admits several parallel programmes, which federation consolidates as plural.

Four domains carry that commitment. Material is the stable referent — artefact, building, site. Observations attach to a digital surrogate at any pixel; structure accumulates from gestures. Digital is what the act produces or invokes — frames, annotation shapes, recordings, session clouds. Those files sit beside the vocabulary, in one package, with the same identifiers as the graph. Knowledge is what is asserted — titles, measures, template fields, vocabulary. Activity is the hinge of the model: it binds the other three. In situ and ex situ qualify the mode of the gesture.

Two symmetric pivots gather the facts. ObservedFeature is a fact documented from the world by an instrumented human act. DerivedFeature is a fact computed from the resulting graph. They recompose in one RDF graph, so a title, a measure or a contour walks back to the gesture that produced it. The OWL 2 vocabulary is multi-typed to PROV-O, CIDOC CRM, schema.org, SKOS and SOSA/SSN, with published bridges, so the same export is readable in those ecosystems.

How activity is situated at three scales

Activity declines into three documentary scales: macro — the programme, the campaign; meso — the session, a field or desk window, with a trajectory and shared resources; micro — the act, the punctual gesture (a frame, a region, a measure, a dictation). Who, why, what, when, where, how and with what qualify that activity at each grain. At macro they are reified as documentation nodes; What and Where are shared with the material referent. At meso they live as a time window, a trajectory or a scene. At micro they are literals on the gesture itself. The same seven questions therefore run through the three scales; the material referent stays one entity.

Time runs through those scales. When qualifies the campaign, the session window and the punctual act. Several meso-sessions sit in one GraphSnapshot. Several agents can annotate one surrogate; the graph keeps each trajectory.

How a graph is produced

dür.air, a mobile digital fieldbook for persistent augmented reality, instantiates the TBox. Minutes of annotation on a tablet yield one spatial and semantic graph, exported as JSON-LD and Turtle with a GraphSnapshot envelope. Each gesture becomes a micro-activity, an observed feature, the media it weaves, the knowledge it states, and — when a deterministic method applies — one or more derived features. Other engines attach at the published pivots. The files those pivots point at — frames, shapes, recordings, clouds — are specified as Folio, the package companion to this model. The Viewer reads that package in the browser.

Similarity buckets are how the graph rises in grain over named subgraphs. An institutional node projects each observation onto four orthogonal axes — spatial, morphology, appearance, semantic — and writes a bucket node when at least two observations fall in the same slot.

The same second pivot also reads a named activity. A StatisticalProfile summarises a programme or a session. A CoverageMatrix records which of the seven documentary axes the campaign filled. A ComparativeReport sets two GraphSnapshots side by side — two objects, two visits, two institutions. That is the collection grain. The public exports already carry those nodes.

Model, Folio, Viewer

Three parts of one observation. The model names it. Folio is the package a producer writes: the files, the poses, the annotations, and the graph. The Viewer opens that package in the browser. A geometric layer can carry both the file to archive and process again, and the declination the browser streams. The source stays in the package. The rendition is what appears.

Activity is the hinge of the model: it binds material, digital and knowledge. The photograph is the hinge of the package: a generic sensor record, the interface of an in situ acquisition, of an ex situ reconstruction, and of any 2D viewpoint in a 3D scene. A stroke, a measure and a voice note are drawn on that image, and the pose carries them onto whatever geometry is loaded. The two hinges are two joints of the same gesture.

The same identifier string runs through the three columns. It is the segment of the IRI, the file name in the package, and the annotation the Viewer selects.

The frame named viewer-z-up is the coordinate system the scene is drawn in. The MiKaDo Viewer is the application that reads a Folio. The dür.air viewer is the field application already running: it reads the bundle dür.air projects, and it draws in that same frame. Documentary, operatory and reflexive are filters of the graph. A geometric layer is the point cloud, the mesh or the gaussian.