How syeddaniyalg/geodialect-hypothesis works
GeoDialect Hypothesis is a reproducible research artifact, not a packaged application: researchers prepare dialect datasets in notebooks, run experiments for three hypotheses, and analyze results for the paper. The reusable H1/H2/H3 implementations and their shared types are the main internal runtime layer; README documentation says H2 and H3 build on H1’s reusable computation state. The README also describes ablations, signal-regime analysis, and figure generation. Notebook internals were not sampled, so links involving those stages reflect documented workflow rather than inspected calls.
Main entry points: native Native experiments · h1.py H1 predictive signal · _types.py Inputs and results · ablations Ablation experiments
Components 17
Research workflow
- Dataset materials
datasets - Dataset registry
configs/datasets.py - Dataset preprocessing
notebooks/processing - Experiment defaults
configs/hypothesis_defaults.py
Hypothesis experiments
- Native experiments
notebooks/hypothesis_experiments/native - Ablation experiments
notebooks/hypothesis_experiments/ablations
Analysis and reporting
- Experiment outputs
results/hypothesis - Signal-regime analysis
notebooks/analysis - Figure generation
notebooks/post-processing - Study paper
paper
Reusable experiment code
- H1 predictive signal
src/geodialect/hypothesis_tests/h1.py - H2 field behavior
src/geodialect/hypothesis_tests/h2.py - H3 exclusion
src/geodialect/hypothesis_tests/h3.py - Inputs and results
src/geodialect/hypothesis_tests/_types.py - Cell execution helpers
src/geodialect/hypothesis_tests/_engine.py - Metrics and inference
src/geodialect/hypothesis_tests/metrics.py
Other components
- Researcher
Connections 26
- Researcher → Dataset preprocessing: prepares data
README.md - Dataset materials → Dataset preprocessing: guides preparation
README.md - Dataset registry → Dataset preprocessing: centralizes paths
configs/datasets.py - Researcher → Native experiments: runs experiments
README.md - Researcher → Ablation experiments: runs experiments
README.md - Experiment defaults → Native experiments: supplies settings
configs/hypothesis_defaults.py - Experiment defaults → Ablation experiments: supplies settings
configs/hypothesis_defaults.py - Native experiments → H1 predictive signal: runs H1
README.md - Native experiments → H2 field behavior: runs H2
README.md - Native experiments → H3 exclusion: runs H3
README.md - Ablation experiments → H1 predictive signal: evaluates hypotheses
README.md - H1 predictive signal → H2 field behavior: provides reusable state
src/geodialect/hypothesis_tests/h2.py - H1 predictive signal → H3 exclusion: provides reusable state
src/geodialect/hypothesis_tests/h3.py - Native experiments → Experiment outputs: saves results
README.md - Ablation experiments → Experiment outputs: saves results
README.md - Experiment outputs → Signal-regime analysis: supports analysis
README.md - Experiment outputs → Figure generation: supports figures
README.md - Signal-regime analysis → Study paper: informs findings
README.md - Figure generation → Study paper: produces figures
README.md - Inputs and results → H1 predictive signal: defines inputs
src/geodialect/hypothesis_tests/h1.py - Inputs and results → H2 field behavior: defines inputs
src/geodialect/hypothesis_tests/h2.py - Inputs and results → H3 exclusion: defines inputs
src/geodialect/hypothesis_tests/h3.py - Cell execution helpers → H1 predictive signal: provides helpers
src/geodialect/hypothesis_tests/h1.py - Cell execution helpers → H2 field behavior: provides helpers
src/geodialect/hypothesis_tests/h2.py - Cell execution helpers → H3 exclusion: provides helpers
src/geodialect/hypothesis_tests/h3.py - Metrics and inference → Inputs and results: uses result types
src/geodialect/hypothesis_tests/metrics.py