v2.0.4: Internationalize codebase to English

What's Changed

Internationalization

• All docstrings, comments, and variable names translated to English
• Bilingual docs structure (docs/en/, docs/pt-BR/) with language switcher
• English README with PT-BR version in docs/pt-BR/
• METHODOLOGY, BIOACOUSTIC_GLOSSARY, COLAB_SETUP translated to English

VMI Label Rename

• M1_DENSO → M1_HEAVY
• M1_LIGEIRO → M1_LIGHT
• M2_REFORCADO → M2_REINFORCED
• M2_LIGEIRO → M2_LIGHT

Outputs

• Report files renamed: analysis_report.md, vmi_analysis.md, llm_report.md
• All plots and reports regenerated with English labels

Full Changelog: v2.0.3...v2.0.4

v2.0.3 - VMI Analysis & Spectral Features

What's New

VMI (Vocal Mechanism Index) Integration

• Added VMI terms and scale to LLM report SYSTEM_PROMPT
• LLM prompt now includes VMI distribution statistics
• New plots: vmi_analysis.png, vmi_scatter.png

Spectral Features in Reports

• Alpha Ratio, H1-H2, Spectral Tilt now included in LLM prompt
• Global means displayed in data summary

Documentation

• Integrated AGNOSTIC.md content into METODOLOGIA.md
• Added Features → VMI mapping table
• Added theoretical notes on CPPS/H1-H2 limitations

Data Management

• Scientific evidence (CSV, PNG, MD) now versioned
• Removed data/processed/* from .gitignore
• Increased pre-commit large file limit to 6MB

Other Changes

• Updated exploratory.py report template (GMM instead of threshold)
• Updated COLAB_SETUP.md with new downloads
• Deleted redundant AGNOSTIC.md

v2.0.1 - RobustScaler

Changes

• fix: Use RobustScaler for clustering normalization
  • Replace manual z-score normalization with sklearn's RobustScaler
  • Uses median/IQR instead of mean/std, more robust to outliers

Full Changelog

v2.0.0...v2.0.1

v2.0.0 - VMI (Vocal Mechanism Index)

🎤 VMI (Vocal Mechanism Index)

Este release introduz o VMI - uma métrica contínua (0-1) para classificação de mecanismo laríngeo.

✨ Novidades

• VMI Contínuo: Métrica 0-1 para classificação de mecanismo laríngeo
• Features Espectrais: Alpha Ratio, H1-H2, Spectral Tilt, CPPS per-frame
• VMIWeights dataclass e VMILabel enum (M1_DENSO → M2_LIGEIRO)
• Suavização Temporal: Métodos median/mean/exponential
• XGBoost Regressor: Treinamento para VMI adaptativo
• Visualizações: VMI scatter plots e timeline

🧪 Qualidade

• Suite de testes completa (59 testes passando)
• Pre-commit hooks configurados (ruff, pytest-quick)
• pytest-cov para cobertura em CI

📚 Documentação

• METODOLOGIA.md atualizado
• Glossário expandido
• COLAB_SETUP documentado

---

Full Changelog: https://github.com/arthurbrenno/vocal-analysis/compare/v1.0.0...v2.0.0