---
license: apache-2.0
datasets:
- Universal-NER/Pile-NER-type
language:
- en
metrics:
- f1
library_name: transformers
pipeline_tag: text2text-generation
---

<p align="center"><h2 align="center">Rethinking Negative Instances for Generative Named Entity Recognition</h2></p>

# Model Card for GNER-T5-large

<!-- Provide a quick summary of what the model is/does. -->

We introduce GNER, a **G**enerative **N**amed **E**ntity **R**ecognition framework, which demonstrates enhanced zero-shot capabilities across unseen entity domains. Experiments on two representative generative models, i.e., LLaMA and Flan-T5, show that the integration of negative instances into the training process yields substantial performance enhancements. The resulting models, GNER-LLaMA and GNER-T5, outperform state-of-the-art (SoTA) approaches by a large margin, achieving improvements of 8 and 11 points in $F_1$ score, respectively. Code and models are publicly available.

* 💻 Code: [https://github.com/yyDing1/GNER/](https://github.com/yyDing1/GNER/)
* 📖 Paper: [Rethinking Negative Instances for Generative Named Entity Recognition](https://arxiv.org/abs/2402.16602)
* 💾 Models in the 🤗 HuggingFace Hub: [GNER-Models](https://huggingface.co./collections/dyyyyyyyy/gner-65dda2cb96c6e35c814dea56)
* 🧪 Reproduction Materials: [Reproduction Materials](https://drive.google.com/drive/folders/1m2FqDgItEbSoeUVo-i18AwMvBcNkZD46?usp=drive_link)
* 🎨 Example Jupyter Notebooks: [GNER Notebook](https://github.com/yyDing1/GNER/blob/main/notebook.ipynb)

<p align="center">
<img src="https://github.com/yyDing1/GNER/raw/main/assets/zero_shot_results.png">
</p>

## PreTrained Models

We release five GNER models based on LLaMA (7B) and Flan-T5 (base, large, xl and xxl).

| Model         | # Params | Zero-shot Average $F_1$ | Supervised Average $F_1$ |          🤗 HuggingFace<br />Download Link          |
| ------------- | -------: | :----------------------: | :-----------------------: | :-------------------------------------------------: |
| GNER-LLaMA    |       7B |           66.1           |           86.09           | [link](https://huggingface.co./dyyyyyyyy/GNER-LLaMA-7B) |
| GNER-T5-base  |     248M |           59.5           |           83.21           | [link](https://huggingface.co./dyyyyyyyy/GNER-T5-base) |
| GNER-T5-large |     783M |           63.5           |           85.45           | [link](https://huggingface.co./dyyyyyyyy/GNER-T5-large) |
| GNER-T5-xl    |       3B |           66.1           |           85.94           |  [link](https://huggingface.co./dyyyyyyyy/GNER-T5-xl)  |
| GNER-T5-xxl   |      11B |           69.1           |           86.15           |  [link](https://huggingface.co./dyyyyyyyy/GNER-T5-xxl)  |

## Demo usage

You should install the dependencies:
```bash
pip install torch datasets deepspeed accelerate transformers protobuf
```

Please check out [Example Jupyter Notebooks](https://github.com/yyDing1/GNER/blob/main/notebook.ipynb) for guidance on utilizing GNER models.

A simple inference example is as follows:

Below is an example using `GNER-T5`
```python
>>> import torch
>>> from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
>>> tokenizer = AutoTokenizer.from_pretrained("dyyyyyyyy/GNER-T5-xxl")
>>> model = AutoModelForSeq2SeqLM.from_pretrained("dyyyyyyyy/GNER-T5-xxl", torch_dtype=torch.bfloat16).cuda()
>>> model = model.eval()
>>> instruction_template = "Please analyze the sentence provided, identifying the type of entity for each word on a token-by-token basis.\nOutput format is: word_1(label_1), word_2(label_2), ...\nWe'll use the BIO-format to label the entities, where:\n1. B- (Begin) indicates the start of a named entity.\n2. I- (Inside) is used for words within a named entity but are not the first word.\n3. O (Outside) denotes words that are not part of a named entity.\n"
>>> sentence = "did george clooney make a musical in the 1980s"
>>> entity_labels = ["genre", "rating", "review", "plot", "song", "average ratings", "director", "character", "trailer", "year", "actor", "title"]
>>> instruction = f"{instruction_template}\nUse the specific entity tags: {', '.join(entity_labels)} and O.\nSentence: {sentence}"
>>> inputs = tokenizer(instruction, return_tensors="pt").to("cuda")
>>> outputs = model.generate(**inputs, max_new_tokens=640)
>>> response = tokenizer.decode(outputs[0], skip_special_tokens=True)
>>> print(response)
"did(O) george(B-actor) clooney(I-actor) make(O) a(O) musical(B-genre) in(O) the(O) 1980s(B-year)"
```

## Citation

```bibtex
@misc{ding2024rethinking,
      title={Rethinking Negative Instances for Generative Named Entity Recognition}, 
      author={Yuyang Ding and Juntao Li and Pinzheng Wang and Zecheng Tang and Bowen Yan and Min Zhang},
      year={2024},
      eprint={2402.16602},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}
```