Batuka0901/mongolian_news_classifier

Mongolian News Headline Classifier

This model is a fine-tuned version of bert-base-multilingual-cased designed to classify Mongolian news headlines into 9 different categories.

Categories

1. Байгал орчин (Environment)
2. Боловсрол (Education)
3. Спорт (Sports)
4. Технологи (Technology)
5. Улс төр (Politics)
6. Урлаг соёл (Arts & Culture)
7. Хууль (Law)
8. Эдийн засаг (Economy)
9. Эрүүл мэнд (Health)

Training Metrics & Results

The model was trained on a custom Mongolian news dataset for 3 epochs. According to the training logs (trainer_state.json), the top performance metrics are:

• Validation Accuracy: 87.11%
• Validation Loss: 0.4071
• Final Training Loss: 0.3286
• Global Steps: 1596

Training Logs Summary

Step	Training Loss	Validation Loss	Validation Accuracy
500	0.5081	0.4627	85.52%
1000	0.4102	0.4344	86.17%
1500	0.3299	0.4112	87.08%
1596	-	0.4071	87.11%

Usage

You can use this model easily with the Hugging Face pipeline:

from transformers import pipeline

# Load the classifier
classifier = pipeline("text-classification", model="Batuka0901/mongolian_news_classifier")

# Predict
headline = "Өнөөдөр хөлбөмбөгийн тэмцээнд Монголын шигшээ баг хожлоо."
result = classifier(headline)

print(result)
# Output: [{'label': 'спорт', 'score': 0.98...}]