Version 0.2.0 Release Notes
Version 0.2.0 adds evidence-first code-switch auditing for short, noisy, mixed-language text.
Included
audit_code_switchingPython API.low-resource-nlp auditCLI command.- Token-level language routes with original character offsets.
- Accepted language spans for mixed text.
- Language-mix counts and code-switch ratio.
- Warnings for weak evidence, ambiguity, low supported-language coverage, and high code-switching.
- Scope notes explaining how the toolkit fits alongside broader African NLP and language-identification work.
Published Artefacts
Scope
This release is not a state-of-the-art African language identifier, a broad language-coverage system, a gold-standard annotation tool, or a substitute for native-speaker review and task-specific evaluation. Routing output should not be treated as identity, ethnicity, nationality, or language-community inference.