lingcore

Abstract linguistic structures for Python.

lingcore provides notation-agnostic building blocks for working with linguistic data: values, features, feature systems, segments, segment inventories, and feature matrices. It has no hardcoded knowledge of IPA, X-SAMPA, or any particular linguistic framework. Consumers bring their own notation and their own feature tables; lingcore provides the structure.

Why

Most phonological or phonetic tooling in Python is either tightly bound to a specific notation (IPA, X-SAMPA) or a specific feature framework (PanPhon, Hayes, PHOIBLE), or it reimplements the same three or four data classes in every project. lingcore factors out the structure so that notation-specific packages can be built on top of a shared, stable foundation.

If you are building a package for IPA, a package for sign-language phonology, or a package for a constructed language you invented last weekend, lingcore gives you the containers without imposing assumptions.

Install

pip install lingcore

Optional extras:

• lingcore[pandas] — for FeatureMatrix.to_dataframe().
• lingcore[dev] — for running the test suite.

Quick example

from lingcore import (
    Feature, FeatureSystem, Segment, SegmentInventory, Value,
    feature_matrix,
)

# Define a feature system.
system = FeatureSystem(
    name="toy",
    version="0.1.0",
    features=(Feature("voice"), Feature("nasal")),
)

# Define an inventory that uses it.
inv = SegmentInventory(
    id="toy",
    types=("consonant",),
    classifiers={"consonant": ("manner", "place")},
)

# Add segments.
inv.add(Segment(
    symbol="p",
    type="consonant",
    classification={"manner": Value("stop"), "place": Value("bilabial")},
    values={"voice": Value("-"), "nasal": Value("-")},
))
inv.add(Segment(
    symbol="m",
    type="consonant",
    classification={"manner": Value("nasal"), "place": Value("bilabial")},
    values={"voice": Value("+"), "nasal": Value("+")},
))

# Materialize a feature matrix.
m = feature_matrix(inv, system)
print(m.distinguishing_features(0, 1))
# ['voice', 'nasal']

What lingcore is not

• Not a tokenizer. It does not know how to split an IPA string into phones. That is the job of a notation-specific package built on top.
• Not a feature table. It does not ship PanPhon or Hayes features. Consumers provide feature tables; lingcore only provides the containers.
• Not an I/O library. It does not read or write YAML, JSON, or CSV. Serialization is a consumer concern.
• Not a full linguistic framework. It has no notion of phonemes, allophony, phonotaxis, syllables, or prosody beyond what is expressed as features. Those are future work, potentially in a sibling package.

Status

Alpha. The API is small and, we hope, stable, but it may change before 1.0. See docs/summary.md for a detailed implementation status.

Documentation

• docs/summary.md — what this project is and where it is going.
• docs/concepts.md — the mental model.
• docs/tutorial.md — a worked example.
• docs/api.md — reference.
• docs/design.md — design decisions and their rationale.

License

GPL-3.0-or-later. See LICENSE.