Role: Developer Org: Open Science Laboratory (OpenSciLab) Distribution: Python package, on GitHub and PyPI
Summary
Converts digits across numeral systems — English, Persian, Hindi, Arabic-Indic, Bengali, and others. Auto-detects mixed formats and converts only the numbers, leaving surrounding text untouched, which makes multilingual and localized data handling straightforward.
The useful part is that second behaviour: XNum finds the numbers inside mixed text and rewrites only those, so a sentence that mixes scripts comes back with its numerals normalized and everything else exactly as it was. Anyone who has cleaned a Persian text by hand knows why the library exists.
It belongs to the natural-language and LLM group of OpenSciLab tools.
Links
Related
naab — Farsi Text Corpus — the other Persian-language data effort. ToCount — Lightweight Token Estimator — sibling text utility in the same OpenSciLab NLP group. Open Source Philosophy — why these tools get written and released at all.