Apppublic

ismatsamadov/azerbaijani-ner-demo

sourceHugging Facemitupdated 6mo agoView on Hugging Face

0likes

App README

Azerbaijani Named Entity Recognition Demo

A Gradio demo for state-of-the-art Named Entity Recognition (NER) in the Azerbaijani language, powered by a fine-tuned XLM-RoBERTa model. Recognizes 25 entity types including persons, locations, organizations, dates, and more.

Live Demo: IsmatS/azerbaijani-ner-demo

Model: IsmatS/xlm-roberta-az-ner

How to Use

1.Enter Azerbaijani text in the input field
2.Click Submit
3.View recognized entities grouped by type
4.Try the built-in example sentences

Model Performance

Metric	Value
Precision	0.76
Recall	0.74
F1 Score	0.75

Entity-Level F1 Scores

Entity	F1
Person	0.85
Location	0.80
Organization	0.69
Percentage	0.81
Product	0.84
Date	0.50

Entity Types Recognized

Entity Type	Description	Example
Person	Names of people	İlham Əliyev
Location	Geographical locations	Bakı, Salyan
Organization	Companies, institutions	Pasha Sığorta
Date	Date expressions	2014-cu il
Time	Time expressions	15:30
Money	Monetary values	500 manat
Percentage	Percentage values	25%
Facility	Buildings, infrastructure	Heydər Əliyev Mərkəzi
Product	Products, goods	iPhone
Government	Government entities	Azərbaycan Respublikası
Law	Legal documents	Konstitusiya
Language	Names of languages	Azərbaycan dili
Event	Events and occurrences	COP29
Art	Artworks	Leyli və Məcnun
Nationality	Nationalities	Azərbaycanlı
Position	Job titles	prezident, nazir

Example Sentences

2014-cu ildə Azərbaycan Respublikasının prezidenti İlham Əliyev Salyanda olub.

Bakı şəhərində Azərbaycan Respublikasının prezidenti İlham Əliyev.

Shahla Khuduyeva və Pasha Sığorta şirkəti haqqında məlumat.

Technical Details

Property	Value
Base Model	XLM-RoBERTa Base (125M parameters)
Fine-tuned on	LocalDoc/azerbaijani-ner-dataset
Training Epochs	8
Batch Size	64
Optimizer	AdamW
GPU	Google Colab Pro+ A100

Quick Start (API)

python

from transformers import AutoTokenizer, AutoModelForTokenClassification, pipeline

model_name = "IsmatS/xlm-roberta-az-ner"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForTokenClassification.from_pretrained(model_name)

ner = pipeline("ner", model=model, tokenizer=tokenizer, aggregation_strategy="simple")

text = "2014-cu ildə Azərbaycan Respublikasının prezidenti İlham Əliyev Salyanda olub."
for entity in ner(text):
    print(f"[{entity['entity_group']}] {entity['word']} ({entity['score']:.2f})")

Related Resources

Citation

bibtex

@model{ismats_az_ner_2024,
  title={XLM-RoBERTa Azerbaijani NER Model},
  author={Ismat Samadov},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/IsmatS/xlm-roberta-az-ner}
}

License

MIT License