Examples¶

Runnable example scripts are in the examples/ directory.

Prerequisites¶

The scripts live in the repository rather than the installed wheel. Clone the repository, enter it, and install the package:

git clone https://github.com/jemsbhai/corpusgen.git
cd corpusgen
python -m pip install -e .

The per-example prerequisites are:

Example	eSpeak NG	PHOIBLE data
01 — inventory exploration	No	Yes
02 — sentence selection	Yes	No
03 — corpus evaluation	Yes	Yes

For Examples 02 and 03, install eSpeak NG using the installation guide. For Examples 01 and 03, download the pinned, checksum-verified PHOIBLE snapshot once:

python -c "from corpusgen.inventory import PhoibleDataset; PhoibleDataset().download()"

01 — Explore a Phoneme Inventory¶

File: examples/01_explore_inventory.py

Load a PHOIBLE phoneme inventory, inspect segments, and query by distinctive features.

python examples/01_explore_inventory.py

What it demonstrates:

Loading inventories with get_inventory()
Listing consonants, vowels, and tones
Querying segments by single or multiple distinctive features
Comparing inventory sizes across languages

02 — Select Sentences for Maximal Coverage¶

File: examples/02_select_sentences.py

Use the greedy selection algorithm to pick a minimal subset of sentences that maximizes phoneme coverage.

python examples/02_select_sentences.py

What it demonstrates:

Selecting sentences with select_sentences()
Inspecting SelectionResult fields (coverage, indices, timing)
Comparing phoneme-level vs diphone-level coverage

03 — Evaluate a Corpus¶

File: examples/03_evaluate_corpus.py

Evaluate an existing corpus against a PHOIBLE target inventory and inspect the full report.

python examples/03_evaluate_corpus.py

What it demonstrates:

Evaluating with evaluate() using auto-discovered and PHOIBLE inventories
Reading distribution metrics (entropy, JSD, PCD)
Reading text quality metrics (sentence length, type-token ratio)
Per-sentence coverage breakdown