Bot Misparim is a script developed for Hebrew Wikipedia for catching common grammar mistakes related to numbers.

In Hebrew one very common mistake is using the wrong gender (שני ילדות; שתי ילדים).

It isn't easy task to parse a sentence and understand to what word the number in the sentence is related to - this bot does it based on simple regex heuristics^* rules. The bot uses hspell project for analyzing words and classify their gender.

^* You are welcome to challenge this approach with more advanced NLP/ML approaches (CFG/PCFGs, LSTM/RLU etc) for catching more grammar errors.

• pywikibot - python framework to access wikipedia
• HspellPy - a python wrapper for hspell

python misparim.py -xml:XML_DUMP

You can get dump from http://dumps.wikimedia.org/

Advances usage:

python misparim.py -xml:XML_DUMP -fix

Semi-automatic bot for fixing the suspected errors. (use with CAUTION)

The bot is heavily based on heuristics so don't always trust it!

Here are some possible mistakes:

• שבע may be
  • זקן ושבע ימים
  • שבע מרורים
• בני תשע + (males)