Arabic NER Model

• Github repo
• NER BIO tagging model based on GigaBERTv4.
• ACE2005 Training data: English + Arabic
• NER tags including: PER, VEH, GPE, WEA, ORG, LOC, FAC

Hyperparameters

• learning_rate=2e-5
• num_train_epochs=10
• weight_decay=0.01

ACE2005 Evaluation results (F1)

How to use

>>> from transformers import pipeline, AutoModelForTokenClassification, AutoTokenizer

>>> ner_model = AutoModelForTokenClassification.from_pretrained("ychenNLP/arabic-ner-ace")
>>> ner_tokenizer = AutoTokenizer.from_pretrained("ychenNLP/arabic-ner-ace")
>>> ner_pip = pipeline("ner", model=ner_model, tokenizer=ner_tokenizer, grouped_entities=True)

>>> output = ner_pip('Protests break out across the US after Supreme Court overturns.')
>>> print(output)
[{'entity_group': 'GPE', 'score': 0.9979881, 'word': 'us', 'start': 30, 'end': 32}, {'entity_group': 'ORG', 'score': 0.99898684, 'word': 'supreme court', 'start': 39, 'end': 52}]

>>> output = ner_pip('قال وزير العدل التركي بكير بوزداغ إن أنقرة تريد 12 مشتبهاً بهم من فنلندا و 21 من السويد')
>>> print(output)
[{'entity_group': 'PER', 'score': 0.9996214, 'word': 'وزير', 'start': 4, 'end': 8}, {'entity_group': 'ORG', 'score': 0.9952383, 'word': 'العدل', 'start': 9, 'end': 14}, {'entity_group': 'GPE', 'score': 0.9996675, 'word': 'التركي', 'start': 15, 'end': 21}, {'entity_group': 'PER', 'score': 0.9978992, 'word': 'بكير بوزداغ', 'start': 22, 'end': 33}, {'entity_group': 'GPE', 'score': 0.9997154, 'word': 'انقرة', 'start': 37, 'end': 42}, {'entity_group': 'PER', 'score': 0.9946885, 'word': 'مشتبها بهم', 'start': 51, 'end': 62}, {'entity_group': 'GPE', 'score': 0.99967396, 'word': 'فنلندا', 'start': 66, 'end': 72}, {'entity_group': 'PER', 'score': 0.99694425, 'word': '21', 'start': 75, 'end': 77}, {'entity_group': 'GPE', 'score': 0.99963355, 'word': 'السويد', 'start': 81, 'end': 87}]

BibTeX entry and citation info

@inproceedings{lan2020gigabert,
  author     = {Lan, Wuwei and Chen, Yang and Xu, Wei and Ritter, Alan},
    title      = {Giga{BERT}: Zero-shot Transfer Learning from {E}nglish to {A}rabic},
    booktitle  = {Proceedings of The 2020 Conference on Empirical Methods on Natural Language Processing (EMNLP)},
    year       = {2020}
  }