IndicBERTv2 is fine-tuned on Sanskrit SampurNER dataset for Fine-grained Named Entity Recognition. It is created using the EaMaTa framework, utilizing the Few-NERD dataset.

SampurNER Paper: SampurNER in AAAI-2026
SampurNER Dataset: prachuryyaIITG/SampurNER

The tagset of Few-NERD is a fine-grained tagset. The fine to coarse level mapping of the tags are as follows:

Location : GPE, Body of Water, Island, Mountain, Park, Road/Transit, Other
Person : Actor, Artist/Author, Athlete, Director, Politician, Scholar, Soldier, Other
ORG : Company, Education, Government, Media, Political Party, Religion, Sports League, Show Organization, Other
Building : Airport, Hospital, Hotel, Library, Restaurant, Sports Facility, Theater, Other
Art : Music, Film, Written Art, Broadcast, Painting, Other
Product : Airplane, Car, Food, Game, Ship, Software, Train, Weapon, Other
Event : Attack, Election, Natural Disaster, Protest, Sports Event, Other
Misc : Astronomy, Award, Biology, Chemistry, Currency, Disease, Educational Degree, God, Language, Law, Living Thing, Medical

Model performance:

Precision: 61.48
Recall: 65.19
F1: 63.27

Training Parameters:

Epochs: 6
Optimizer: AdamW
Learning Rate: 5e-5
Weight Decay: 0.01
Batch Size: 64

Contributors

Prachuryya Kaushik
Prof. Ashish Anand

It is part of the AWED-PIPER ecosystem: Paper | Agent for FgNER | Web App for FgNER | Agent for PII Protection | Web App for PII Protection

Sample Usage

The AWED-FiNER agentic tool can be used to interact with expert models trained using this framework. Below is an example:

pip install smolagents gradio_client

python

from tool import AWEDFiNERTool

tool = AWEDFiNERTool(
    space_id="prachuryyaIITG/AWED-FiNER"
)

result = tool.forward(
    text="Jude Bellingham joined Real Madrid in 2023.",
    language="English"
)

print(result)

Citation

If you use this model, please cite the following papers:

bibtex

@inproceedings{kaushik2026sampurner,
      title={SampurNER: Fine-Grained Named Entity Recognition Dataset for 22 Indian Languages},
      volume={40},
      url={https://ojs.aaai.org/index.php/AAAI/article/view/40405},
      DOI={10.1609/aaai.v40i37.40405},
      number={37},
      journal={Proceedings of the AAAI Conference on Artificial Intelligence},
      author={Kaushik, Prachuryya and Anand, Ashish},
      year={2026},
      month={Mar.},
      pages={31410-31418}
}

@misc{kaushik2026awedpiperagentswebapplications,
      title={AWED-PIPER: Agents, Web Applications & Expert Detectors for Personally Identifiable Information Protection & Fine-grained Named Entity Recognition across 36 languages for 6.6 Billion Speakers}, 
      author={Prachuryya Kaushik and Ashish Anand},
      year={2026},
      eprint={2601.10161},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2601.10161}, 
}

@inproceedings{ding-etal-2021-nerd,
    title = "Few-{NERD}: A Few-shot Named Entity Recognition Dataset",
    author = "Ding, Ning and Xu, Guangwei and Chen, Yulin and Wang, Xiaobin and Han, Xu and Xie, Pengjun and Zheng, Haitao and Liu, Zhiyuan",
    booktitle = "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)",
    month = aug,
    year = "2021",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2021.acl-long.248",
    doi = "10.18653/v1/2021.acl-long.248",
    pages = "3198--3213",
}

prachuryyaIITG/SampurNER_Sanskrit_IndicBERTv2

Model performance:

Training Parameters:

Contributors

Sample Usage

Citation

prachuryyaIITG/
SampurNER_Sanskrit_IndicBERTv2