Training-Free Cross-Lingual Dysarthria Severity Assessment via Phonological Subspace Analysis in Self-Supervised Speech Representations B. Muller, A.A. Ortiz Barrañón, and L. Roberts The Scott-Morgan Foundation / Tecnológico de Monterrey Contents Analysis pipeline: HuBERT phone-level embedding extraction, d' computation, 12-metric phonological profile Phone feature configurations for 6 languages (English, Dutch, Spanish, French, Mandarin, Italian) Statistical analysis scripts (bootstrap CIs, FDR correction, LOCO, meta-analysis, ROC) MFA alignment scripts Figure generation scripts Requirements Python 3.12 PyTorch 2.10 HuggingFace Transformers 4.57 Montreal Forced Aligner 3.3 See requirements.txt for pinned versions. Citation If you use this code, please cite: @article{muller2026phonological, title={Training-Free Cross-Lingual Dysarthria Severity Assessment via Phonological Subspace Analysis in Self-Supervised Speech Representations}, author={Muller, Bernard and Ortiz Barrañón, Antonio Armando and Roberts, LaVonne}, year={2026} } License Apache 2.0