NuNER: Entity Recognition Encoder Pre-training via LLM-Annotated Data

NuNER is the family of SOTA Foundation and Zero-shot for Entity Recognition

Foundation models:

HuggingFace

• NuNER 2.0: numind/NuNER-v2.0 (MIT) - the most powerful English NER model
• NuNER multilingual numind/NuNER-multilingual-v0.1 (MIT) - the most powerful multilingual NER model
• NuNER 1.0 numind/NuNER-v1.0 (MIT) - the main model from our paper
• NuNER BERT numind/NuNER-BERT-v1.0(MIT) - the model used in the TadNER section of our paper

Zero-shot models:

HuggingFace

• GLiNER NuNerZero span: numind/NuNER_Zero-span (MIT) - +4.5% more powerful GLiNER Large v2.1
• NuNerZero: numind/NuNER_Zero (MIT) - +3% more powerful GLiNER Large v2.1, better suitable to detect multi-word entities
• NuNerZero 4k context: numind/NuNER_Zero-4k (MIT) - 4k-long-context NuNerZero

The last 2 models are word-level GLiNER models with unlimited entity length supported

Citation

@misc{bogdanov2024nuner,
      title={NuNER: Entity Recognition Encoder Pre-training via LLM-Annotated Data}, 
      author={Sergei Bogdanov and Alexandre Constantin and Timothée Bernard and Benoit Crabbé and Etienne Bernard},
      year={2024},
      eprint={2402.15343},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}