Add Australian Curriculum v9 (F–6) as a standards source, plus candidate crosswalk

[curiousape-ai]

This PR adds Australia to the taxonomy's curriculum-standards sources, following the repo's existing conventions.

What's included

1. data/curriculum-standards.json — a new acara-v9 source (8th curriculum).
All 614 Foundation–Year 6 content descriptions of the Australian Curriculum v9, with full text: code, subject, learning area, strand, year levels, description. Validated against schema/curriculum-standards.schema.json. ACARA publishes under CC BY 4.0 (attribution, no share-alike), so this is a 🟢 full-text source — licensing details added to PROVENANCE.md. The dataset is parsed from ACARA's official DOCX downloads with a reproducible, tested parser: https://github.com/curiousape-ai/acara-v9-f6-json.

2. crosswalks/acara-v9-f6-candidates.json — machine-generated candidate alignments (clearly separated from your curated standards links).
Every one of the 1,590 topics classified against the AC9 codes: 1,007 mapped with confidence bands (233 high / 461 medium / 313 low), 445 no-F6-match, 120 beyond-F6, 18 general-capabilities. Method: sentence-transformer similarity (all-MiniLM-L6-v2) with TF-IDF boost, constrained by subject compatibility and year overlap. v0.1, frozen, unreviewed — these are candidates for human review, not curated alignments, which is why they live in a separate file and topics.json is untouched. A known limitation is disclosed in the file's meta.limitations: v0.1 was matched against 475 of the 614 content descriptions (a parsing gap excluded sub-strand content); no_f6_match is therefore an upper bound and a v0.2 re-match is planned.

3. PROVENANCE.md — ACARA licensing section (CC BY 4.0, required attribution notice), inserted alongside the other full-text source.

Why

The README says the taxonomy aims to align to national curriculum standards — this adds the Australian mapping so any consumer gets it natively. The no-match analysis may also be useful signal for you: the 445 unmatched topics cluster where Australian sequencing differs from UK/US practice (AI literacy has no F–6 home at all; ancient civilisations are deferred to Year 7; grammar is named less explicitly).

Notes for maintainers

• If you'd prefer the candidates file in a different location/format — or dropped from the PR entirely, keeping only the standards source — happy to restructure; the standards addition stands on its own.
• No NSW (state-level) syllabus text is included anywhere; that source is not openly licensed.
• I'm happy to maintain the Australian source going forward, including re-derivation when ACARA revises v9 and the planned v0.2 of the candidates.

---

Australian Curriculum material © ACARA, licensed under CC BY 4.0.