README.md

---
license: apache-2.0
base_model:
  - indobenchmark/indobert-base-p1
pipeline_tag: text-classification
tags:
  - indobert
  - intent-classification
  - jkn-kis
  - bpjs
  - bahasa-indonesia
  - healthcare
  - nlp
language:
  - id
library_name: transformers
model_type: bert
---

# 🩺 IndoBERT for Intent Classification on JKN-KIS Data

This model is a fine-tuned version of [`indobenchmark/indobert-base-p1`](https://huggingface.co/indobenchmark/indobert-base-p1), adapted for **intent classification** on question data related to the **Jaminan Kesehatan Nasional – Kartu Indonesia Sehat (JKN-KIS)** program.

The model is trained to recognize user intent from common questions asked in the context of **BPJS Kesehatan** services, such as registration, benefits, payment, eligibility, and more.

---

## 🚀 How to Use

Make sure to load your model and tokenizer, and also the label encoder (`le`) if you're using `LabelEncoder` from `sklearn`.

```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch
import joblib  # only needed if you use sklearn LabelEncoder

# Load model and tokenizer
model_name = "vinapatri/intent-classification-jkn-kis"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# Load the label encoder used during training
le = joblib.load("label_encoder.pkl")

def predict_intent(text):
    inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True)

    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    model.to(device)
    inputs = {k: v.to(device) for k, v in inputs.items()}

    with torch.no_grad():
        outputs = model(**inputs)

    logits = outputs.logits
    predicted_class_id = logits.argmax().item()
    tag = le.inverse_transform([predicted_class_id])[0]
    return tag

# Example
text = "Apa tata cara memperoleh surat keterangan tidak mampu untuk BPJS?"
predicted_intent = predict_intent(text)
print(f"Intent: {predicted_intent}")