Featured Open Source Work
Hazm Library Contribution
I contributed to Hazm, an open-source Python library for Persian natural language processing, by adding model-driven modules, pretrained resources, and improvements that make the library easier to use across platforms.
Hazm 0.8
- Implemented word embedding and sentence embedding modules.
- Added pretrained FastText word embeddings and sentence embedding models.
Hazm 0.9
- Improved Windows compatibility by moving from Wapiti to Python-crfsuite.
- Added pretrained Chunker and POS tagger models.
- Implemented embedding training workflows and an EmbedRank keyword extraction example.
Hazm 0.9.2
- Added pretrained dependency parser models.
- Improved project test coverage.
PyTorch / NLP
Rumour Detection
A PyTorch classification project for experimenting with rumour detection models, preprocessing options, and training configurations.
Persian NLP
Persian POS Tagger
A comparative implementation of Persian part-of-speech tagging with neural and classical approaches.
Recommendation
Book Recommendation System
A recommendation engine for a book donation platform using text embeddings, cosine similarity, and PCA.
Keyphrase Extraction
Persian EmbedRank
A Persian keyphrase extraction system adapted from the EmbedRank approach and integrated with Hazm-based models.
Recommendation
Clothes Recommendation System
A preference-based recommendation system that represents user taste as vectors and matches items with similarity and clustering methods.
Dataset / Detection
Persian AI-Generated Text Detection
A dataset and model exploration project for detecting AI-generated Persian text.