urdu-nlp 0.1.0

A lightweight, pure-Python NLP library for Urdu language processing

urdu-nlp

A lightweight, pure-Python NLP library for Urdu language processing. No deep learning required — just install and go.

Urdu is spoken by 230+ million people, yet has almost no usable NLP tooling on PyPI. urdu-nlp fills that gap with tokenization, stop word removal, normalization, stemming, transliteration, and sentence boundary detection.

Installation

pip install urdu-nlp

Or install from source:

git clone https://github.com/imabd645/urdu-nlp.git
cd urdu-nlp
pip install -e .

Quick Start

1. Tokenization

from urdu_nlp import tokenize

tokenize("میں اسکول جاتا ہوں")
# → ["میں", "اسکول", "جاتا", "ہوں"]

2. Stop Word Removal

from urdu_nlp import remove_stopwords

remove_stopwords(["میں", "اسکول", "جاتا", "ہوں"])
# → ["اسکول", "جاتا"]

3. Normalization

from urdu_nlp import normalize

normalize("ﻛﺮﻧﺎ")   # Arabic form
# → "کرنا"          # Urdu form

4. Stemming

from urdu_nlp import stem

stem("کتابوں")
# → "کتاب"

5. Roman Urdu → Urdu Script

from urdu_nlp import roman_to_urdu

roman_to_urdu("mein school jata hoon")
# → "میں اسکول جاتا ہوں"

6. Sentence Boundary Detection

from urdu_nlp import sent_tokenize

sent_tokenize("یہ پہلا جملہ ہے۔ یہ دوسرا ہے۔")
# → ["یہ پہلا جملہ ہے۔", "یہ دوسرا ہے۔"]

API Reference

Function	Description
tokenize(text)	Split Urdu text into word tokens
sent_tokenize(text)	Split text into sentences
remove_stopwords(tokens)	Remove common Urdu stop words from a token list
normalize(text)	Normalize Arabic/Urdu character variants and whitespace
stem(word)	Strip common Urdu suffixes to get root form
roman_to_urdu(text)	Transliterate Roman Urdu to Urdu script

Dependencies

• regex — Unicode-aware pattern matching (the only dependency)

License

MIT