language
multilingual af am ar as az be bg bn br bs ca cs cy da de el en eo es et eu fa fi fr fy ga gd gl gu ha he hi hr hu hy id is it ja jv ka kk km kn ko ku ky la lo lt lv mg mk ml mn mr ms my ne nl no om or pa pl ps pt ro ru sa sd si sk sl so sq sr su sv sw ta te th tl tr ug uk ur uz vi xh yi zh

xlm-roberta-large-finetuned-conll03-english

Table of Contents

1. Model Details
2. Uses
3. Bias, Risks, and Limitations
4. Training
5. Evaluation
6. Environmental Impact
7. Technical Specifications
8. Citation
9. Model Card Authors
10. How To Get Started With the Model

Model Details

Model Description

The XLM-RoBERTa model was proposed in Unsupervised Cross-lingual Representation Learning at Scale by Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer and Veselin Stoyanov. It is based on Facebook's RoBERTa model released in 2019. It is a large multi-lingual language model, trained on 2.5TB of filtered CommonCrawl data. This model is XLM-RoBERTa-large fine-tuned with the conll2003 dataset in English.

• Developed by: See associated paper
• Model type: Multi-lingual language model
• Language(s) (NLP) or Countries (images): XLM-RoBERTa is a multilingual model trained on 100 different languages; see GitHub Repo for full list; model is fine-tuned on a dataset in English
• License: More information needed
• Related Models: RoBERTa, XLM
  • Parent Model: XLM-RoBERTa-large
• Resources for more information: -GitHub Repo -Associated Paper

Uses

Direct Use

The model is a language model. The model can be used for token classification, a natural language understanding task in which a label is assigned to some tokens in a text.

Downstream Use

Potential downstream use cases include Named Entity Recognition (NER) and Part-of-Speech (PoS) tagging. To learn more about token classification and other potential downstream use cases, see the Hugging Face token classification docs.

Out-of-Scope Use

The model should not be used to intentionally create hostile or alienating environments for people.

Bias, Risks, and Limitations

CONTENT WARNING: Readers should be made aware that language generated by this model may be disturbing or offensive to some and may propagate historical and current stereotypes.

Significant research has explored bias and fairness issues with language models (see, e.g., Sheng et al. (2021) and Bender et al. (2021)). In the context of tasks relevant to this model, Mishra et al. (2020) explore social biases in NER systems for English and find that there is systematic bias in existing NER systems in that they fail to identify named entities from different demographic groups (though this paper did not look at BERT). For example, using a sample sentence from Mishra et al. (2020):

>>> from transformers import pipeline
>>> tokenizer = AutoTokenizer.from_pretrained("xlm-roberta-large-finetuned-conll03-english")
>>> model = AutoModelForTokenClassification.from_pretrained("xlm-roberta-large-finetuned-conll03-english")
>>> classifier = pipeline("ner", model=model, tokenizer=tokenizer)
>>> classifier("Alya told Jasmine that Andrew could pay with cash..")
[{'end': 2,
  'entity': 'I-PER',
  'index': 1,
  'score': 0.9997861,
  'start': 0,
  'word': '▁Al'},
 {'end': 4,
  'entity': 'I-PER',
  'index': 2,
  'score': 0.9998591,
  'start': 2,
  'word': 'ya'},
 {'end': 16,
  'entity': 'I-PER',
  'index': 4,
  'score': 0.99995816,
  'start': 10,
  'word': '▁Jasmin'},
 {'end': 17,
  'entity': 'I-PER',
  'index': 5,
  'score': 0.9999584,
  'start': 16,
  'word': 'e'},
 {'end': 29,
  'entity': 'I-PER',
  'index': 7,
  'score': 0.99998057,
  'start': 23,
  'word': '▁Andrew'}]

Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model.

Training

See the following resources for training data and training procedure details:

• XLM-RoBERTa-large model card
• CoNLL-2003 data card
• Associated paper

Evaluation

See the associated paper for evaluation details.

Environmental Impact

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

• Hardware Type: 500 32GB Nvidia V100 GPUs (from the associated paper)
• Hours used: More information needed
• Cloud Provider: More information needed
• Compute Region: More information needed
• Carbon Emitted: More information needed

Technical Specifications

See the associated paper for further details.

Citation

BibTeX:

@article{conneau2019unsupervised,
  title={Unsupervised Cross-lingual Representation Learning at Scale},
  author={Conneau, Alexis and Khandelwal, Kartikay and Goyal, Naman and Chaudhary, Vishrav and Wenzek, Guillaume and Guzm{\'a}n, Francisco and Grave, Edouard and Ott, Myle and Zettlemoyer, Luke and Stoyanov, Veselin},
  journal={arXiv preprint arXiv:1911.02116},
  year={2019}
}

APA:

• Conneau, A., Khandelwal, K., Goyal, N., Chaudhary, V., Wenzek, G., Guzmán, F., ... & Stoyanov, V. (2019). Unsupervised cross-lingual representation learning at scale. arXiv preprint arXiv:1911.02116.

Model Card Authors

This model card was written by the team at Hugging Face.

How to Get Started with the Model

Use the code below to get started with the model. You can use this model directly within a pipeline for NER.

Click to expand

>>> from transformers import AutoTokenizer, AutoModelForTokenClassification
>>> from transformers import pipeline
>>> tokenizer = AutoTokenizer.from_pretrained("xlm-roberta-large-finetuned-conll03-english")
>>> model = AutoModelForTokenClassification.from_pretrained("xlm-roberta-large-finetuned-conll03-english")
>>> classifier = pipeline("ner", model=model, tokenizer=tokenizer)
>>> classifier("Hello I'm Omar and I live in Zürich.")

[{'end': 14,
  'entity': 'I-PER',
  'index': 5,
  'score': 0.9999175,
  'start': 10,
  'word': '▁Omar'},
 {'end': 35,
  'entity': 'I-LOC',
  'index': 10,
  'score': 0.9999906,
  'start': 29,
  'word': '▁Zürich'}]