cagrickr/turkish-voice-banking-nlp

🚀 Live Demo Available

👉 https://huggingface.co/spaces/cagrickr/banking-voice-demo

---

🇹🇷 Turkish Voice Banking NLP Model

This model is an intent classification system designed to understand Turkish banking commands.

🎯 Purpose:

• Analyze user voice or text commands
• Identify the correct banking intent
• Enable voice-based banking assistant applications

---

🚀 Features

• Turkish language support
• Banking-focused intent classification
• Multi-command support (e.g. "show my balance and list recent transactions")
• Real-time inference capability
• Supports both voice and text input
• Integrated demo application

---

🧠 Supported Intents

• bakiye_sorgulama
• hesap_hareketleri
• iban_hesap_bilgisi
• eft_havale_fast
• transfer_talimat_limit
• kredi_karti_borc
• kredi_karti_limit_ekstre
• kart_bloke_guvenlik
• sanal_kart
• fatura_odeme
• otomatik_odeme
• kredi_basvuru_detay
• doviz_kur
• altin_hesabi
• atm_sube_bulma
• sifre_hesap_guvenligi

---

🧪 Example Usage

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model_name = "cagrickr/turkish-voice-banking-nlp"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

text = "Show my balance"

inputs = tokenizer(text, return_tensors="pt")

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

pred = torch.argmax(outputs.logits, dim=1).item()

print(pred)

---

🎤 Demo

You can try the live demo here:

👉 https://huggingface.co/spaces/cagrickr/banking-voice-demo

---

🏗️ Model Details

• Model type: BERT-based classification model
• Framework: PyTorch
• Library: Hugging Face Transformers
• Language: Turkish
• Use cases: Voice Banking, Chatbots, Fintech applications

---

📊 Training Data

• Banking-related commands (manually created and expanded)
• 50 base commands → 300+ variations
• Data augmentation applied
• Designed to simulate real user scenarios

---

📈 Performance (Approximate)

• Accuracy: ~92%
• F1 Score: ~90%

---