master-thesis 1.0.0

About the Project

The source code, data, and experiment results for the Master Thesis "Unsupervised Segmentation of Gregorian Chant Melodies for Exploring Chant Modality" created at Faculty of Mathematics and Physics, Charles University.

Key Features

1. Nested Hierarchical Pitman-Yor Language Model
   • implementation in Cython
2. Chant Segmentations
   • generated by the trained NHPYLM model based on the mode extension (eight independent NHPYLM submodels)
3. Evaluation Scores
   • measuring chant segmentation quality, modality relations, segmentation analysis, modality analysis
4. Best Practices + Experiment Results
   • jupyter notebooks
5. Filtered Datasets
   • Antiphons+Responsories