hazm

Why we included this project

Persian text is messy in ways that generic NLP libraries often ignore, and Hazm exists specifically to handle that mess. It normalizes Farsi input by fixing diacritics, ZWNJ, and orthographic variants, then provides tokenization, lemmatization, part-of-speech tagging, dependency parsing, and word or sentence embeddings. Pretrained models load automatically from Hugging Face, so you can get from raw text to structured linguistic features with a few lines of Python instead of training your own. The corpus-reading utilities are a handy bonus for researchers who want ready-made access to standard Persian datasets. For anyone doing serious NLP work in Persian, this is a solid, well-documented place to start.