maai-project/maai-theme-sl

MAAI Theme Classifier (Single Label)

This model is a fine-tuned distilroberta-base that predicts the primary theme of a climate-related article or paragraph.

Task

Given an input text (sentence, paragraph, or article), the model predicts exactly one theme label.

Labels

The full label set and mappings are provided in label_map.json.

Model Details

• Architecture: DistilRoBERTa
• Task type: Single-label text classification
• Base model: distilroberta-base

Training

• Dataset: Final_20k_Articles_for_MaaiBert.csv
• Epochs: 4
• Train batch size: 16
• Evaluation batch size: 32

Evaluation

Evaluation results on the test set are available in test_metrics.json.

Usage

from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="maai-project/maai-theme-sl"
)

classifier("New policies aim to reduce carbon emissions and introduce carbon taxes.")

Notes

• This model predicts one label only per input.
• For best results, use complete sentences or paragraphs.