N'Ko Brain Scanner — Comprehensive Project Handoff

An ASR (Automatic Speech Recognition) system for N'Ko script — a phonetically transparent writing system used by ~30M Manding-language speakers in West Africa. The core research question: **does N'Ko's phonetic transparency give it a measurable architectural advantage over Latin script in ASR?** Current verified answer: **the N'Ko trajectory model is strong enough to anchor the project baseline.** The fully archived reproduction on the current `290,596`-pair corpus snapshot achieves **20.57% CER**. Earlier 8-way N'Ko/Latin comparison numbers remain useful historical internal evidence, but their full artifact bundle is not yet restored on this machine and should be treated as provisional. | Property | Value | |----------|-------| | Architecture | UnifiedCTCHead (46.8M params) | | Input | Whisper large-v3 encoder features (1280-dim) | | Decoder | 6-layer Transformer with trajectory bias injection | | Training data | 290,596 pairs (232,476 train / 29,060 val / 29,060 test) | | Seed | 42 (deterministic split) | | Epochs trained | 46 (best checkpoint at epoch 38) | | Best val loss | 0.6359 | | Checkpoint | `results/paper4_reproduction_35205256/best.pt` | | Results JSON | `results/paper4_reproduction_35205256/results.json` | | Inference script | `asr/transcribe_nko.py` | Artifacts are also synced to `Mac5:/Volumes/HD1/tar_297k_clean/paper4_reproduction_35205256/`. ### Trajectory Bias Mechanism The model injects pen-stroke trajectory scalars (7-dim: velocity, curvature, acceleration, etc.) into transformer attention as position-dependent bias. This exploits N'Ko's bijective grapheme-phoneme mapping — each character encodes exactly one phoneme, and the trajectory captures how that character is physically written.