xlm-roberta-large-ner

Named Entity Recognition (NER) Model for Uzbek Language

About the Model

This model is designed for Named Entity Recognition (NER) in Uzbek text. The model can identify various categories of named entities, including persons, places, organizations, dates, and more. This model is based on the XLM-RoBERTa large architecture.

Note!!!

The model is trained on the NEWS dataset and primarily has high accuracy for identifying NER in NEWS texts.

Categories

The model can identify the following NER categories:

• LOC (Location names)
• ORG (Organization names)
• PERSON (Person names)
• DATE (Date expressions)
• MONEY (Monetary amounts)
• PERCENT (Percentage values)
• QUANTITY (Quantities)
• TIME (Time expressions)
• PRODUCT (Product names)
• EVENT (Event names)
• WORK_OF_ART (Work of art titles)
• LANGUAGE (Language names)
• CARDINAL (Cardinal numbers)
• ORDINAL (Ordinal numbers)
• NORP (Nationalities or religious/political groups)
• FACILITY (Facility names)
• LAW (Laws or regulations)
• GPE (Countries, cities, states)

Examples

To demonstrate how the model works, here are a few examples:

from transformers import pipeline, AutoTokenizer, AutoModelForTokenClassification

model_name_or_path = "risqaliyevds/xlm-roberta-large-ner"
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)
model = AutoModelForTokenClassification.from_pretrained(model_name_or_path).to("cuda")

nlp = pipeline("ner", model=model, tokenizer=tokenizer)

text = "Shavkat Mirziyoyev Rossiyada rasmiy safarda bo'ldi."
ner = nlp(text)

for entity in ner:
    print(entity)

Example text: "Shavkat Mirziyoyev Rossiyada rasmiy safarda bo'ldi."

Results:

[{'entity': 'B-PERSON', 'score': 0.88995147, 'index': 1, 'word': '▁Shavkat', 'start': 0, 'end': 7},
 {'entity': 'I-PERSON', 'score': 0.980681, 'index': 2, 'word': '▁Mirziyoyev', 'start': 8, 'end': 18},
 {'entity': 'B-GPE', 'score': 0.8208886, 'index': 3, 'word': '▁Rossiya', 'start': 19, 'end': 26}]

Loading and Using the Model

To download and use the model from the Hugging Face platform, you can use the following code:

from transformers import pipeline, AutoTokenizer, AutoModelForTokenClassification

model_name_or_path = "risqaliyevds/xlm-roberta-large-ner"

tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)
model = AutoModelForTokenClassification.from_pretrained(model_name_or_path).to("cuda")

nlp = pipeline("ner", model=model, tokenizer=tokenizer)

Contact

If you have any questions or need more information, please contact us. LinkedIn: Riskaliev Murad

License

This model is provided as open source and is available for free use by all users.

Conclusion

The NER model for the Uzbek language is effective in identifying various named entities in texts. The high accuracy and wide range of categories make it useful for academic research, document analysis, and many other fields.