Skip to content

Polish Whisper Normalizer

Polish text normalizer for Whisper ASR output – preserves ąćęłńóśźż, normalizes numbers, time, currency, dates, fractions, half and declensions via Morfeusz2. Diacritic-less ASR (czterdziesci, piec) is handled automatically.

from polish_whisper_normalizer import PolishTextNormalizer

n = PolishTextNormalizer()
n("Spotkanie dwudziestego pierwszego maja o piętnastej trzydzieści.")
# → "spotkanie 21.05 o 15:30"
n("trzysta czterdziesci osiem")
# → "348"

Quickstart

uv sync                 # install + morfeusz2
uv run pytest -q        # 500+ tests
uv run mkdocs serve     # docs at http://127.0.0.1:8000

Features

  • Diacritics – ąćęłńóśźż preserved, diacritic-less fallback via Morfeusz
  • Numbers – cardinal/ordinal with declensions via PolishLemmatizer (Morfeusz2)
  • Time – HH:MM, wpół do, za 15, po 15, o 5:00, 5:00 rano, od 5:00 do 6:00
  • Currency/percent/fractions/half – 5 zł, 5%, 1/3, 0.5 litra (fraction via ordinal lemmas)
  • Dates – 5 maja → 05.05, piątego maja 2026 → 05.05.2026, conditional maja vs Maja
  • Geographic guard – na północ stays, jest północ → 0:00

See API for details.

Pipeline

lower → brackets/ignore → r.→roku → sentence periods/ellipsis → time → decimal ,→. → numeral boundaries → remove_symbols → numbers → months (conditional) → dates → ordinal-dot cleanup (PolishTextNormalizer in text.py)

Architecture

src/polish_whisper_normalizer/
  basic.py        # Whisper basic, diacritics-aware
  lemmatizer.py   # Morfeusz2 wrapper (analyse/generate)
  utils.py        # ASCII variants
  numbers.py      # cardinals/ordinals/currency
  time.py         # HH:MM
  text.py         # full pipeline
  polish.py       # re-export shim

Development

uv sync --group dev
uv run ruff check src tests
uv run ruff format --check src tests
uv run mypy src
uv run pytest --cov --cov-report=term-missing
uv build
mkdocs serve