kmjun3203/bert-based-uncased-argnews4-v01

BERT Fine-tuned on AG News for 4-Class Text Classification

Model Description

This model fine-tunes the bert-base-uncased model on the AG News dataset for text classification.
It classifies English news articles into one of four categories:

Label	English Category	Korean Label
0	World	세계뉴스
1	Sports	스포츠
2	Business	비즈니스
3	Science/Technology	과학/기술

Intended Uses & Limitations

Intended Uses

• News article categorization
• Content tagging for media organizations
• NLP education and experimentation

Limitations

• Trained only on English text → may not generalize to other languages
• May contain biases from AG News dataset
• Requires further evaluation for production or fairness tasks

Training Details

• Base model: bert-base-uncased
• Dataset: AG News
• Task: Text classification (4 classes)
• Tokenizer: BertTokenizer (uncased, max length 128)
• Optimizer: AdamW
• Loss function: CrossEntropyLoss
• Batch size: 16
• Epochs: 3
• Learning rate: 2e-5

Evaluation

Metric	Score
Accuracy	82.5%

Usage Example

from transformers import AutoTokenizer, AutoModelForSequenceClassification import torch

model_name = "your-username/bert-agnews-korean-labels" tokenizer = AutoTokenizer.from_pretrained(model_name) model = AutoModelForSequenceClassification.from_pretrained(model_name)

text = "Apple announces the new iPhone at their annual event." inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True) outputs = model(**inputs) predicted_class = torch.argmax(outputs.logits).item()

label_map = { 0: "세계뉴스", 1: "스포츠", 2: "비즈니스", 3: "과학/기술" }

print("Predicted label:", label_map[predicted_class])

Ethical Considerations

• May reflect label bias inherent in AG News dataset.
• Should not be used for misinformation or surveillance.
• Verify performance before high-impact use cases.

Citation

If you use this model, please cite the base paper: @article{devlin2018bert, title={BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding}, author={Devlin, Jacob and Chang, Ming-Wei and Lee, Kenton and Toutanova, Kristina}, journal={arXiv preprint arXiv:1810.04805}, year={2018} }

---