README.md

metadata

license: cc-by-4.0
language:
  - de
tags:
  - legal-nlp
  - annotation
  - interpretive-canons
  - label-studio
pretty_name: Interpretive Canons — Raw Label Studio Export (pre-merge snapshot)

Classifying Interpretive Canons — Raw Label Studio Export (pre-merge snapshot)

Raw, pre-merge Label Studio annotation export (project 46, snapshot 2026-07-25). Two fields needed to reconstruct the konkretes_gesetz (statutory-reference) subtask survive only in this pre-merge form: the per-reading determinate-content judgment bestimmt_moeglicheDeutung, which the postprocessing merge drops entirely, and the list of concrete provisions, which the merge flattens into a single unsplittable string (keeping only the first entry). Join to the benchmark by Label Studio region id.

Why this snapshot exists

The published benchmark (felix453/interpretive-canons-instances) is built from a merged/postprocessed export. That merge step is lossy in two ways that matter for reconstructing the statutory-reference subtask:

• bestimmt_moeglicheDeutung is dropped. A valid konkretes_gesetz negative is a reading that reached the criterion (cleared the abstract and self-advanced gates) and carries a recorded bestimmt answer (Yes/No). A span that never reached the criterion, or was left unanswered, is not a negative. That judgment exists only here.
• The concrete-provision list is flattened. Label Studio stores a multi-line TextArea answer as a list, one entry per provision; the merge joins them into one string and keeps only the first. The intact list survives only here.

Contents

• raw-labelstudio-export-project-46-2026-07-25.json — the raw export, one task per decision (28 decisions), with per-region annotations under annotations[].result.

Companion datasets

• felix453/interpretive-canons-instances — instance-level benchmark (the eight subtasks, train/validation/test).
• felix453/interpretive-canons-decisions — decision-level release.