Low-Resource NLP Toolkit 0.3.0

Version 0.3.0 adds selective language routing and the first measured benchmark for the package.

The routing API can reject decisions below configurable confidence or score-margin thresholds while retaining the underlying evidence. Coverage-aware evaluation reports accuracy and coverage for accepted routes, with per-language results.

The AfriSenti benchmark pins the official source revision and file hashes for five languages: Hausa, Igbo, Nigerian Pidgin, Swahili and Yoruba. A policy selected on development data reached 89.65% accuracy at 74.03% coverage on 18,402 test examples. The paired orthography test found that removing diacritics reduced overall accuracy from 74.75% to 69.87%.

Source tweets are downloaded locally and are not included in this repository or its reports.

Method and per-language results, including limitations: https://oyinkanchekwas.github.io/low-resource-nlp-toolkit/benchmark/

low-resource-nlp-toolkit 0.2.1

Version 0.2.1 corrects the package information displayed by PyPI and the citation metadata shipped with the project. The Python API and command-line behaviour are unchanged from 0.2.0.

low-resource-nlp-toolkit 0.2.0

Adds an evidence-first code-switch audit for short multilingual text. The audit reports token routes, original offsets, accepted spans, language-mix counts, a code-switch ratio and warnings for weak or mixed evidence. The package remains dependency-light and works as a corpus-triage layer. Broad African language identification requires dedicated trained models and wider evaluation.

v0.1.0

Seed release of the Low-Resource NLP Toolkit. This release provides dependency-light utilities for text normalisation, language routing, emotion label harmonisation, evaluation, datasets, and the CLI. It deliberately avoids private datasets, committed model weights, API-backed demos, and claims that routing identifies a person or community. Verified before release: local quality gate and 24 tests passed; GitHub Actions passed on main; isolated source and wheel builds passed; twine check passed; the built wheel installed cleanly and the CLI smoke tests passed.