CorpusMind v0.1.20 — Full Reference Corpora (BNC Baby, BAWE, Leipzig)
Phase 2: Full Reference Corpora
New Full Corpus Downloads
- BNC Baby (4M words, BNC User Licence) — 4 subcorpora: academic, fiction, news, spoken
- BAWE (6.5M words, CC BY-NC-SA 3.0) — academic English with discipline metadata
- Leipzig English News 10K (CC BY 4.0) — full sentences with NLP pipeline
- Leipzig Arabic News 10K (CC BY 4.0) — Arabic news with NLP pipeline
New Endpoint
POST /reference-corpora/{name}/download-full — downloads archive, extracts text, creates Corpus row, ingests through NLP pipeline, tags documents with genre metadata
UI Improvements
- Full Corpus vs Frequency List badge in BundledReferences
- Auto-detection of format (full_corpus calls download-full endpoint)
Verified
- ruff: 0 errors
- pytest: 24/24 passed
- typecheck: clean
- build: clean
Full Changelog: v0.1.19...v0.1.20