The taxonomy workflow

Instruments detect when things sound; deciding what they are — in Schaeffer's typo-morphology or Schafer's soundscape functions — is a listening judgment. ambiscape splits the work accordingly:

ambiscape draft <folder>      # machine proposes
# ... you listen and edit ...
ambiscape taxonomy <folder>   # machine renders

1. Draft (annotations.draft.json)

From the cached features, draft pre-fills:

• steady-state keynote candidates — level regimes found by change-point detection with a fixed per-regime reference (median of the regime's first two minutes) and a two-minute confirmation window, so transients don't split regimes but machine on/offs do;
• detected events, each with listening hints: clock time, exceedance, level, azimuth/elevation, diffuseness — and, with the [ml] extra, AudioSet tag suggestions from PANNs (treat as suggestions to confirm by ear, not ground truth).

2. Annotate (annotations.json)

For each object you set:

Field	Values
kind	keynote | signal | soundmark | figure
mass	tonic | tonic-complex | complex | noise
facture	impulse | iteration | sustained | unlimited
soundmark (optional)	community | dwelling
source (optional)	anthropophony | biophony | geophony
spans / events	times as "[D ]HH:MM:SS" (D = days after day 0)

plus an optional states list for lo-fi spans (e.g., a masking drone). The full schema is documented in the taxonomy module; worked examples live in the Intercontinental database's Haarlem and Berlin session folders.

3. Render

taxonomy produces two figures:

• Schaeffer map — every object on the facture × mass plane, colored by its Schafer function; a corpus's structure in one glance (keynotes crowd the sustained/unlimited columns, signals the impulse/iteration columns).
• Schafer timeline — one lane per object on the real session clock: keynote bars, signal/soundmark event markers, lo-fi states shaded, gap-aware panels for multi-take sessions. Makes claims like "the community soundmark is audible only in hi-fi windows" visible directly.