Core Concepts
Talonic turns unstructured documents into schema-validated data with per-cell provenance. These six concepts explain how the platform works under the hood.
The Field Registry
The unified knowledge graph of canonical fields discovered across all your documents. Maturity levels (Core, Proven, Candidate), semantic clustering, and master extraction instructions.
Schemas as Entities
Schemas are versioned entities that define extraction targets. Each field links to the Field Registry and tracks changes via the Schema Graph.
Cases
Document groups connected through shared entities. Cases auto-form from link keys and include evidence chains with AI narration.
Per-Cell Provenance
Every extracted value carries a confidence score, source region, reasoning trace, and resolution method — the confidence gate locks cells at 0.7+.
The Pipeline & Spec Rail
A Spec composes a rail of stages over one schema; a pipeline is one run of that Spec over your documents, with per-cell confidence and provenance.
Document Ontology
A 529-type classification system aligned with DIN SPEC 91491 for consistent document type recognition across languages and industries.
For hands-on usage, see the Extract endpoint, Schemas API, or the authentication guide.