Taxonomy reference

The taxonomy is an author-editable contract at backend/config/scribe/taxonomy.yaml. It defines the entity types the knowledge-synthesis workflow assigns to extracted concepts, and their hierarchy. Editing it changes what Avalon can classify — it is data, not code.

Schema

The file has one top-level key, entity_types, mapping each type name to a definition:

entity_types:
  PRODUCT:
    description: "Financial products and services offered by DNB and Sbanken…"
    parent: null              # top of a hierarchy
    subtypes: [LOAN, INSURANCE, INVESTMENT, CARD, PENSION, ACCOUNT, PAYMENT_SERVICE]
    examples: ["dnb-boliglån", "sbanken-bilforsikring"]

  LOAN:
    description: "Lending products for personal and business use…"
    parent: PRODUCT            # nested under PRODUCT
    subtypes: []
    examples: ["dnb-boliglån", "sbanken-billån", "dnb-byggelån"]

Field	Meaning
description	Human-readable definition; also steers the classifier.
parent	The enclosing type, or null for a root.
subtypes	Child type names (must themselves be defined as entries).
examples	Representative entity slugs of this type.

Hierarchy rules

• Every parent and every entry in subtypes must be a defined entity_types key.
• A root type has parent: null; a leaf has subtypes: [].
• parent/subtypes must agree: if LOAN.parent is PRODUCT, then LOAN should appear in PRODUCT.subtypes.

Entity-slug naming

examples use lowercase, hyphenated slugs prefixed by brand — dnb-… or sbanken-… (dnb-boliglån, sbanken-bilforsikring). The prefix identifies the owning brand; the remainder is the product/entity name in Norwegian.

Where it is used

The Knowledge & Avalon synthesis workflow reads this file to label concepts before consolidation into Avalon.