atha-text-classifier

Model ini adalah fine-tuned indobenchmark/indobert-base-p1 untuk klasifikasi sentimen Bahasa Indonesia 3 kelas.

Label output:

• negative
• neutral
• positive

Training data: https://huggingface.co/datasets/Bangkah/atha-text-dataset

Quick Use (Transformers)

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model_id = "Bangkah/atha-text-classifier"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForSequenceClassification.from_pretrained(model_id)

text = "produk ini bagus dan pengirimannya cepat"
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=128)
with torch.no_grad():
    logits = model(**inputs).logits

probs = torch.softmax(logits, dim=-1)[0]
label_id = int(torch.argmax(probs).item())
label = model.config.id2label[label_id]
score = float(probs[label_id].item())
print({"label": label, "confidence": round(score, 4)})

Limitations

• Dataset training masih sintetis, sehingga metrik tinggi tidak langsung merepresentasikan performa produksi.
• Untuk use-case production, tetap lakukan fine-tuning ulang dengan data real domain aplikasi.

Validation Metrics

• Loss: 0.0004
• Accuracy: 1.0000
• Macro F1: 1.0000

Confusion Matrix

true\pred	negative	neutral	positive
negative	100	0	0
neutral	0	100	0
positive	0	0	100

Classification Report

              precision    recall  f1-score   support

    negative     1.0000    1.0000    1.0000       100
     neutral     1.0000    1.0000    1.0000       100
    positive     1.0000    1.0000    1.0000       100

    accuracy                         1.0000       300
   macro avg     1.0000    1.0000    1.0000       300
weighted avg     1.0000    1.0000    1.0000       300