no-nepali-profanity

A small, dependency-free profanity matcher for English, Romanized (Latin) Nepali and Devanagari Nepali, plus the Hindi slang common in Nepal. Built for moderating user-written text — names, comments, reviews — on Nepali sites, where false positives on real names are more damaging than a missed swear.

Zero runtime dependencies. Python ≥ 3.9. A direct port of the no-nepali-profanity npm package, with the same word lists and matching rules.

Documentation: mukhxadnahunna.com/python

Install

pip install no-nepali-profanity

Usage

from no_nepali_profanity import contains_profanity, find_profanity, tokenize

# boolean check — fastest
contains_profanity("Great teacher!")        # False
contains_profanity("muji")                  # True
contains_profanity("मुजीको कक्षा")           # True  (Devanagari + postposition)

# which words — returns normalised matches as they appeared
find_profanity("f.u.c.k this sh1t")        # ["fuck", "shit"]
find_profanity("f u c k this")              # ["fuck"]
find_profanity("Randip Thapa")              # []
find_profanity("*ss teacher")               # ["*ss"]

# debugging — see what the matcher actually splits into
tokenize("Great teacher!")                  # ["great", "teacher"]

API

Match positions (ProfanityMatch.start, end) are character (code point) indices, unlike the npm package which uses UTF-16 code units.

Censoring

censor("you muji")                                   # "you ****"
censor("F.U.C.K this Sh1t!")                          # "******* this ****!"
censor("you muji", {"mask": "#"})                     # "you ####"
censor("you muji", {"replace": lambda m: "[censored]"})  # "you [censored]"
find_profanity_matches("you muji")                     # [ProfanityMatch(text="muji", normalized="muji", start=4, end=8)]

Check and censor in one pass:

result = check("you muji")
result.has_profanity   # True
result.words           # ["muji"]
result.censor()        # "you ****"

Options

find_profanity("fuck muji मुजी", {"languages": ["romanized"]})      # ["muji"]
contains_profanity("you idiot", {"strictness": "lenient"})         # False
find_profanity("terms and conditions", {"strictness": "strict"})   # ["conditions"]

• languages: any of "english", "romanized", "devanagari". Default: all three.
• strictness: "lenient" (severe words only), "standard" (default, adds milder insults like idiot, murkha) or "strict" (adds the stems rand, cond, kand, lund, which also hit words like Randip and conditions).

What it catches

• Case and Unicode forms: IDIOT, full-width letters.
• Leetspeak: sh1t, @ss (0 1 3 4 5 7 @ $).
• ! for i between letters: sh!t, b!tch. Sentence-final Great teacher! is left alone.
• * for a hidden letter: f*ck, sh*t, and markdown emphasis like *sh*t* still reads as the word.
• Stretched letters: fuuuuck, for words of 4+ letters.
• Spelled-out letters: f.u.c.k, f u c k.
• Nepali postpositions and plurals glued on: mujiko, randiharu, मुजीको, …हरू.
• Devanagari spelling variants: nukta, chandrabindu vs anusvara, zero-width joiners.
• Stems where no ordinary word starts the same way: fucking, bitches, machiknee.
• Multi-word phrases: chaak ko pwal, pesa garne, sasto manche (Latin and Devanagari).

What it deliberately doesn't

• Short words match exactly, so as, class, assignment and Assam are fine.
• Name collisions: shit is a whole word only, because Shitij / शितिज is a name. Names like मुजी-adjacent Randip / राण्डीप, Putali / पुतली, Asha / आशा are checked in the test suite.
• No caste names, surnames or ordinary words that are only offensive in context (e.g. kami, kukur). A word list can't tell a slur from someone's name; that needs human moderation.
• No judgement of context, sarcasm or meaning. This is a first-pass filter, not a moderator.

Development

python -m venv .venv && .venv/bin/pip install -e ".[test]"
.venv/bin/pytest

The word lists live in src/no_nepali_profanity/lexicon.py, apart from the matching logic in src/no_nepali_profanity/core.py. Native-speaker review of the Nepali lists is the most valuable contribution.

License

MIT