---
datasets:
- l3cube-pune/MahaParaphrase
language:
- mr
license: cc-by-4.0
metrics:
- accuracy
pipeline_tag: text-classification
library_name: transformers
---

# MahaParaphrase-BERT Model

**MahaParaphrase-BERT** is a **MahaBERT** model ([l3cube-pune/marathi-bert-v2](https://huggingface.co/l3cube-pune/marathi-bert-v2)) fine-tuned on the **L3Cube-MahaParaphrase Dataset**, a high-quality Marathi paraphrase detection corpus. The dataset consists of 8,000 sentence pairs annotated as Paraphrase (P) or Non-paraphrase (NP).

This model is trained specifically for Marathi Paraphrase Detection task.

More details on the model, training methodology, and evaluation results can be found in the paper: [MahaParaphrase: A Marathi Paraphrase Detection Corpus and BERT-based Models](https://huggingface.co/papers/2508.17444).

For further resources, including the **L3Cube-MahaParaphrase Dataset** and code, visit the [L3Cube-MahaNLP GitHub repository](https://github.com/l3cube-pune/MarathiNLP).

## Usage

This model can be used with the Hugging Face `transformers` library. For detailed usage examples and more information on the `mahaNLP` library and its integration, please refer to the [L3Cube-MahaNLP GitHub repository](https://github.com/l3cube-pune/MarathiNLP) and its associated [Colab notebook](https://colab.research.google.com/drive/1POx3Bi1cML6-s3Z3u8g8VpqzpoYCyv2q).

## Citing

If you use this model or the associated dataset, please cite the following paper:
```bibtex
@article{mahaparaphrase2025,
  title={MahaParaphrase: A Marathi Paraphrase Detection Corpus and BERT-based Models},
  author={Please refer to the paper for authors},
  journal={arXiv preprint arXiv:2508.17444},
  year={2025},
  url={https://arxiv.org/abs/2508.17444}
}
```