ryulecca/indobert_nadiem

Indonesian Stance Detection Model

Model klasifikasi stance bahasa Indonesia menggunakan fine-tuned IndoBERT.

Model Details

• Base Model: indobenchmark/indobert-base-p1
• Task: Stance Detection (6 kelas)
• Training Data: 5900 sampel
• Validation Data: 677 sampel

Labels

Label	Sentimen
pro_nadiem	Positif
kontra_nadiem	Negatif
kritik_pemerintah	Negatif
kritik_peradilan	Negatif
netral_informasional	Netral
tak_jelas	Netral

Usage

from transformers import pipeline

classifier = pipeline("text-classification", model="ryulecca/indobert_nadiem", return_all_scores=True)
result = classifier("Nadiem Makarim melakukan pekerjaan yang bagus")
print(result)

Evaluation Results

{
  "eval_loss": 1.9400033950805664,
  "eval_accuracy": 0.5878877400295421,
  "eval_recall": 0.48985290643465484,
  "eval_f1": 0.49022874462253047
}