FAC Implementation Scorecard

- Entries: 105 - Unique videos: 20 - N'Ko chars: 12541 - Tone marks: 3316 - Parsed syllables: 4139 - Marked register H+L: 65.8% - Non-contour H+L+M: 99.1% - Contour rising+falling: 0.9% - PASS: unicode_inventory_verified - PASS: tone_prior_current_corpus - PASS: text_baseline_recomputed - PASS: scripts_pass - PASS: contract_tests_pass - PASS: no_stale_claims - BLOCKED: read_speech_available - BLOCKED: checkpoint_inference_verified - PASS: `/opt/homebrew/opt/python@3.14/bin/python3.14 -W ignore tone_lm_baseline.py` - corpus: 105 lessons, 4139 syllables - text-only tone-resolution TDER (avg of 5 splits, lower=better): - majority-class floor : 58.7% - PASS: `/opt/homebrew/opt/python@3.14/bin/python3.14 -W ignore tone_seam_v0.py` - ================================================================ - TONE-SEAM v0 — acoustic tone resolution for N'Ko - ================================================================ - PASS: `/opt/homebrew/opt/python@3.14/bin/python3.14 -W ignore tone_fusion_eval.py --selftest` - [selftest] controlled F0 matching gold tones (acoustic should be near-perfect): - syllables=60 - TEXT-only TDER : 60.0% - PASS: `/opt/homebrew/opt/python@3.14/bin/python3.14 -W ignore h2_pitch_fidelity.py` - H2 pitch-fidelity | RMSE in cents (lower=better) | rate-distortion - corpus: synthetic tonal, 500x6x16, F0 100-250 Hz, contour depth 360 cents - - PASS: `/opt/homebrew/opt/python@3.14/bin/python3.14 -m pytest -q tests` - ...... [100%] - 6 passed in 0.03s - A real read-speech WAV aligned to known toned N'Ko text. - One checkpoint inference pass to measure end-to-end toned CER.