Skip to content

Features Overview

The Norm Editor turns the open-ended task of "reading a law and writing down what it means" into a guided, structured workflow. This section describes the capabilities the editor offers at each stage.

At a glance

Feature What it does
Guided interpretation workflow A five-step stepper that walks the interpreter from task definition to a finished interpretation
The FLINT frame model Fact, Act, and Claim-duty frames with typed roles and subtypes
Source annotation Select sentences from a structured document and highlight fragments to create frames
Boolean constructs Compose preconditions and fact subdivisions with AND / OR / NOT
NLP assistance Machine-learning suggestions for the constituents of an act frame
Frame visualisation View an interpretation as a filterable list or as an interactive network graph
TriplyDB integration & formats Load and save sources and interpretations as RDF, TriG, or JSON

Guided interpretation workflow

The editor is organised as a stepper with five stages. The first three are fully functional; the last two are placeholders for future work.

graph LR
    S1[1 · Define a task] --> S2[2 · Collect sources]
    S2 --> S3[3 · Interpret sources]
    S3 --> S4[4 · Validate interpretations]
    S4 --> S5[5 · Perform task]

    style S1 fill:#4a90e2,color:#fff
    style S2 fill:#4a90e2,color:#fff
    style S3 fill:#4a90e2,color:#fff
    style S4 fill:#dddddd
    style S5 fill:#dddddd
  1. Define a task — record who is doing the interpretation (the editor), a label, and a description. Each task receives its own stable identifier and is linked to exactly one interpretation.
  2. Collect sources — load one or more normative documents and select the sentences that are in scope. Sources can come from a server file, from TriplyDB, or from the local file system.
  3. Interpret sources — the main working area, with the source text on the left and the frames on the right. Here the interpreter highlights fragments, creates and edits frames, assigns roles, and builds boolean preconditions.
  4. Validate interpretations (planned) — reserved for checking the interpretation.
  5. Perform task (planned) — reserved for executing the interpretation.

A persistent banner across the stepper header keeps load and save actions available at every step, so an interpretation can be saved or reopened at any time.

Designed for interpreters, not RDF authors

Every feature is built around the idea that the person doing the work is a legal or policy expert, not a knowledge engineer:

  • Frames are created by highlighting text, never by typing IRIs.
  • Labels for act and claim-duty frames are generated automatically from their roles.
  • The complex RDF serialisation is handled entirely by the conversion services.
  • Comments can be attached to any frame to record interpretation decisions for reviewers.

The pages in this section describe each capability in more depth. For step-by-step instructions, see the User Guide.