MuSTDrifter

MuSTDrifter (Multi-Source Temporal Drifter) is a Python framework for unsupervised multidimensional discourse drift detection.

The framework quantifies how discourse evolves over time through complementary forms of distribution shift analysis, combining:

• Covariate shift detection
• Prior probability shift detection
• Multidimensional discourse representations
• Drift estimation and reporting utilities

MuSTDrifter supports discourse analysis across:

• Semantic dimensions
• Lexical dimensions
• Syntactic content
• Syntactic style
• Thematic distributions

Features

• Multidimensional discourse representations
• Multiple drift metrics
• Parallelized computation
• Permutation-based inference
• Automatic report generation
• Heatmap visualization utilities
• Modular and extensible architecture

Supported Drift Metrics

Semantic Drift

• Maximum Mean Discrepancy (MMD)
• Cosine Drift
• Kolmogorov–Smirnov (KS)

Lexical, Syntactic, and Thematic Drift

• Jensen–Shannon Divergence (JS)
• Kullback–Leibler Divergence (KL)
• Log-Likelihood divergence

Documentation

• Installation
• Quickstart
• API Reference

Citation

TBD