Ministry-of-ICT-Internship-2026/Luganda-Text-Normalizer

# Luganda Text Normalizer A Python library for preprocessing Luganda text for NLP pipelines: cleaning, elision expansion, abbreviation/acronym expansion, currency normalization, date/time expansion, number-to-words expansion, tokenization, dictionary-based spelling standardization, noun/verb morphology stripping, and English–Luganda code-switching detection. > Status: actively developed. See `docs/architecture.md` > for the full design rationale, module inventory, and known limitations. ## Install From the project root, editable install (recommended for development): ```bash pip install -e . # with dev/test dependencies: pip install -e ".[dev]" ``` Requires Python ≥ 3.9. Depends on `nltk>=3.8` (see `requirements.txt`). ## Quick start ```python import normalizer normalizer.normalize("Nasasula UGX 50,000 nga 16/07/2026 essaawa 07:00.") ``` Or use individual stages directly, without the full pipeline: ```python from normalizer import number_to_words, expand_currency_to_words, expand_abbreviations number_to_words(45) # -> "amakumi ana mu ttaano" expand_currency_to_words("Ekiguzibwa $10 nedda.") # -> "Ekiguzibwa ddoola kkumi nedda." expand_abbreviations("URA basoma tax.") # -> "Uganda Revenue Authority basoma tax." ``` ## What it does The pipeline covers ten preprocessing capabilities, each implemented in its own module so you can use just the pieces you need: | # | Capability | Module | Example | |---|---|---|---| | 1 | Unicode normalization | `cleaner` | NFC-normalizes text, fixes stray control characters | | 2 | Lowercasing | `pipeline` | applied after tokenization, before spelling lookup | | 3 | Punctuation cleanup | `cleaner` | standardizes quotes/dashes, collapses repeated punctuation | | 4 | Number expansion | `number_words` | `25` → `amakumi abiri mu ttaano` | | 5 | Date/time expansion | `datetime_expand` | `16/07/2026` → `Lwakutaano, olunaku 16 mu mwezi gwa Julaayi, mu mwaka 2026` | | 6 | Currency normalization | `currency` | …