Decipher
Docs
decipher / docs
v0.1.0 — computational toolkit

Overview

  • Introduction
  • Getting started
  • Core concepts
  • Data products
  • Audio gallery

Tutorials

  • Index
  • Load and listen
  • Detect acoustic units
  • Cluster units into types
  • Build a symbol stream
  • Entropy & mutual information
  • Zipf and Menzerath

Guides

  • The round-trip invariant
  • Spectrograms
  • Tuning segmentation
  • Embeddings (AVES & autoencoder)
  • Repertoire diversity

API reference

  • Index
  • decipher.audio
  • decipher.cluster
  • decipher.stream
  • decipher.featurise
  • decipher.ssl
  • decipher.autoencoder
  • decipher.infotheory
  • decipher.sequence
  • decipher.stats
  • decipher.diversity
  • decipher.catalog

Tutorials

Each tutorial is a working pipeline you can run end-to-end against your own audio.

1. Load and listen

Read a WAV, inspect provenance, write it back. Verify the round-trip invariant.

2. Detect acoustic units

Bandpass, then segment with the energy-envelope detector. Tune thresholds.

3. Cluster units into types

Featurise with mel patches, reduce with PCA, cluster with HDBSCAN.

4. Build a symbol stream

Combine cluster labels and units into a SymbolStream with session boundaries.

5. Entropy and mutual information

Shannon entropy with Miller-Madow correction; bias-corrected MI with a permutation null.

6. Zipf and Menzerath

Test whether the repertoire follows the linguistic-laws predictions.