L3Cube-IndicNLP

The L3Cube's IndicNLP project is an effort to improve NLP resources for Indic languages. We have created monolingual BERT models for 10 Indic languages. We have also released monolingual and multilingual (cross-lingual) Sentence BERT models. These models provide state-of-the-art results on downstream tasks.

Monolingual BERT models for Indic languages

More details about these models can be found in paper

Model	Link
Marathi BERT	model
Hindi BERT	model
Dev BERT (Hindi + Marathi)	model
Kannada BERT	model
Telugu BERT	model
Malayalam BERT	model
Tamil BERT	model
Gujarati BERT	model
Oriya BERT	model
Bengali BERT	model
Punjabi BERT	model
Assamese BERT	model

Indic Sentence BERT models

More details about these models can be found in paper

Similarity Model	Sentence BERT
Marathi Similarity	Marathi SBERT
Hindi Similarity	Hindi SBERT
Kannada Similarity	Kannada SBERT
Telugu Similarity	Telugu SBERT
Malayalam Similarity	Malayalam SBERT
Tamil Similarity	Tamil SBERT
Gujarati Similarity	Gujarati SBERT
Oriya Similarity	Oriya SBERT
Bengali Similarity	Bengali SBERT
Punjabi Similarity	Punjabi SBERT
Indic Similarity (multilingual)	Indic SBERT (multilingual)

License

All the resources are licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. The datasets are released to the community for research purposes only and the group is not responsible for any misuse of these datasets.

Citing

@article{joshi2022l3cube_hind,
  title={L3Cube-HindBERT and DevBERT: Pre-Trained BERT Transformer models for Devanagari based Hindi and Marathi Languages},
  author={Joshi, Raviraj},
  journal={arXiv preprint arXiv:2211.11418},
  year={2022}
}

@article{deode2023l3cube,
  title={L3Cube-IndicSBERT: A simple approach for learning cross-lingual sentence representations using multilingual BERT},
  author={Deode, Samruddhi and Gadre, Janhavi and Kajale, Aditi and Joshi, Ananya and Joshi, Raviraj},
  journal={arXiv preprint arXiv:2304.11434},
  year={2023}
}

Publications

Joshi, Raviraj. "L3Cube-HindBERT and DevBERT: Pre-Trained BERT Transformer models for Devanagari based Hindi and Marathi Languages." arXiv preprint arXiv:2211.11418 (2022).
Deode, Samruddhi, et al. "L3Cube-IndicSBERT: A simple approach for learning cross-lingual sentence representations using multilingual BERT." arXiv preprint arXiv:2304.11434 (2023).
Mirashi, Aishwarya, et al. "L3Cube-IndicNews: News-based Short Text and Long Document Classification Datasets in Indic Languages." arXiv preprint arXiv:2401.02254 (2024).

This project is led by Raviraj Joshi under L3Cube Labs, Pune. For any queries contact ravirajoshi@gmail.com .