AzrilFahmiardi/sdd-emotion-general

Model Card: Emotion General

Model Overview

Model Name: sdd-emotion-general
Base Model: xlm-roberta-base
Task: Emotion classification (6-class)
Language: Indonesian

---

Model Description

Fine-tuned XLM-RoBERTa for detecting emotions in Indonesian text.

Emotions:

• anger (Anger, frustration)
• fear (Fear, anxiety)
• happy (Happiness, joy)
• love (Love, affection)
• neutral (Neutral tone)
• sadness (Sadness, grief)

---

Performance Metrics

Metric	Value
Accuracy (EmoT test)	0.7364
Macro F1	0.7423
Latency (mean)	10.5 ms
Model Size	7.5 GB

---

Usage

Load Model

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model_name = "AzrilFahmiardi/sdd-emotion-general"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = model.to(device)

Inference

def detect_emotion(text: str) -> dict:
    inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=128).to(device)
    
    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs.logits
    
    probabilities = torch.softmax(logits, dim=-1)[0].cpu()
    predicted_class = logits.argmax(-1).item()
    predicted_label = model.config.id2label[predicted_class]
    confidence = probabilities[predicted_class].item()
    
    return {
        "emotion": predicted_label,
        "confidence": confidence
    }

# Example
text = "Saya sangat senang bisa bersama keluarga hari ini!"
result = detect_emotion(text)
print(f"Emotion: {result['emotion']} ({result['confidence']:.2%})")

Output Format

{
  "emotion": "happy",
  "confidence": 0.8934
}

---

Input/Output

Parameter	Type	Example
Input	str	Indonesian text, max 128 tokens
Output	dict	{"emotion": "happy", "confidence": 0.89}