Daily Mirror News Classifier
DistilBERT fine-tuned to classify Sri Lankan English news excerpts into five categories.
Categories
| Label | ID |
|---|---|
| Business | 0 |
| Opinion | 1 |
| Political_gossip | 2 |
| Sports | 3 |
| World_news | 4 |
Usage
from transformers import pipeline
classifier = pipeline("text-classification", model="Dineth0627/daily-mirror-news-classifier_")
result = classifier("The Central Bank raised rates to curb inflation.")
print(result) # [{'label': 'Business', 'score': 0.97}]
Training details
| Setting | Value |
|---|---|
| Base model | distilbert-base-uncased |
| Epochs | 4 (EarlyStopping patience=2) |
| Learning rate | 2e-5 |
| Batch size | 16 |
| Max sequence length | 128 |
| Warmup ratio | 0.10 |
| Weight decay | 0.01 |
| Optimiser | AdamW + linear schedule |
Dataset
~1,015 labelled news excerpts from Daily Mirror Online (Sri Lanka), split 80 / 10 / 10 (train / val / test) with stratification.
- Downloads last month
- 5
Model tree for Dineth0627/daily-mirror-news-classifier_
Base model
distilbert/distilbert-base-uncased