Upload docs/model_cards/finee-dataset-README.md with huggingface_hub

docs/model_cards/finee-dataset-README.md

@@ -0,0 +1,210 @@
+---
+license: apache-2.0
+language:
+- en
+- hi
+- ta
+- te
+- bn
+- kn
+size_categories:
+- 100K<n<1M
+task_categories:
+- token-classification
+- text-generation
+tags:
+- finance
+- banking
+- entity-extraction
+- indian-banking
+- sms
+- synthetic
+pretty_name: FinEE Dataset - Indian Financial Entity Extraction
+---
+
+# FinEE Dataset
+
+<p align="center">
+  <img src="https://img.shields.io/badge/Samples-152K%2B-blue" alt="Samples">
+  <img src="https://img.shields.io/badge/Languages-6-orange" alt="Languages">
+  <img src="https://img.shields.io/badge/Banks-25%2B-green" alt="Banks">
+</p>
+
+## Dataset Description
+
+A comprehensive dataset for training financial entity extraction models on Indian banking messages. Contains 152,000+ samples covering SMS, emails, and transaction notifications from major Indian banks.
+
+### Languages
+
+- English (en) - 86%
+- Hindi (hi) - 3%
+- Tamil (ta) - 3%
+- Telugu (te) - 3%
+- Bengali (bn) - 3%
+- Kannada (kn) - 2%
+
+### Supported Transaction Types
+
+- UPI payments (PhonePe, GPay, Paytm)
+- NEFT/IMPS/RTGS transfers
+- Credit card transactions
+- Debit card transactions
+- EMI payments
+- Refunds and reversals
+- Salary credits
+- Bill payments
+
+### Covered Banks
+
+HDFC, ICICI, SBI, Axis, Kotak, PNB, BOB, Canara, Union, IDBI, IndusInd, Yes Bank, Federal, South Indian, Karur Vysya, and more.
+
+### Covered Merchants
+
+- Food: Swiggy, Zomato, Zepto, BigBasket
+- Shopping: Amazon, Flipkart, Myntra, Meesho
+- Travel: Uber, Ola, IRCTC, MakeMyTrip
+- Investment: Zerodha, Groww, Upstox, Angel One
+- Bills: Airtel, Jio, electricity, gas
+- Entertainment: Netflix, BookMyShow, Hotstar
+
+## Dataset Structure
+
+### Data Fields
+
+```json
+{
+  "input": "HDFC Bank: Rs.2,500 debited from A/c XX1234...",
+  "output": {
+    "amount": 2500.0,
+    "type": "debit",
+    "account": "1234",
+    "bank": "HDFC",
+    "merchant": "Swiggy",
+    "category": "food",
+    "is_p2m": true
+  }
+}
+```
+
+### Instruction Format (ChatML)
+
+```json
+{
+  "messages": [
+    {"role": "system", "content": "You are a financial entity extraction assistant..."},
+    {"role": "user", "content": "Extract financial entities from: ..."},
+    {"role": "assistant", "content": "{\"amount\": 2500.0, ...}"}
+  ]
+}
+```
+
+### Splits
+
+| Split | Samples | Description |
+|-------|---------|-------------|
+| train | 137,267 | Training data |
+| valid | 7,625 | Validation data |
+| test | 7,627 | Test data (held out) |
+
+## Data Sources
+
+1. **Real Data** (2,419 samples)
+   - Anonymized ICICI Bank SMS messages
+   - Manually verified labels
+
+2. **Synthetic Data** (100,000 samples)
+   - Grammar-based generation
+   - Covers all bank templates
+   - Realistic amount distributions
+
+3. **Multilingual Synthetic** (50,100 samples)
+   - Hindi, Tamil, Telugu, Bengali, Kannada
+   - Markov chain for realistic flow
+   - Edge case oversampling
+
+## Usage
+
+### Load with Datasets
+
+```python
+from datasets import load_dataset
+
+dataset = load_dataset("Ranjit0034/finee-dataset")
+
+# Access splits
+train = dataset["train"]
+valid = dataset["valid"]
+test = dataset["test"]
+
+# Iterate
+for example in train:
+    print(example["input"])
+    print(example["output"])
+```
+
+### Load for Fine-tuning
+
+```python
+from datasets import load_dataset
+
+# Load instruction format
+dataset = load_dataset("Ranjit0034/finee-dataset", data_files={
+    "train": "instruction/train.jsonl",
+    "valid": "instruction/valid.jsonl"
+})
+```
+
+## Output Schema
+
+| Field | Type | Description |
+|-------|------|-------------|
+| amount | float | Transaction amount in INR |
+| type | string | "debit" or "credit" |
+| account | string | Last 4 digits of account |
+| bank | string | Bank name |
+| date | string | Transaction date (YYYY-MM-DD) |
+| time | string | Transaction time (HH:MM) |
+| reference | string | UPI/NEFT reference number |
+| merchant | string | Merchant name (P2M) |
+| beneficiary | string | Person name (P2P) |
+| vpa | string | UPI VPA address |
+| category | string | Transaction category |
+| is_p2m | boolean | true if merchant, false if P2P |
+| balance | float | Balance after transaction |
+| status | string | success/failed/pending |
+
+## Categories
+
+- `food` - Restaurants, delivery
+- `grocery` - Supermarkets
+- `shopping` - E-commerce, retail
+- `transport` - Cab, fuel
+- `travel` - Flights, hotels
+- `bills` - Utilities, recharge
+- `entertainment` - Movies, streaming
+- `healthcare` - Medical, pharmacy
+- `investment` - Stocks, mutual funds
+- `transfer` - P2P transfers
+- `salary` - Income
+- `emi` - Loan payments
+
+## Citation
+
+```bibtex
+@dataset{finee_dataset,
+  title={FinEE Dataset: Indian Financial Entity Extraction},
+  author={Ranjit Behera},
+  year={2026},
+  url={https://huggingface.co/datasets/Ranjit0034/finee-dataset}
+}
+```
+
+## License
+
+Apache 2.0
+
+## Related
+
+- 🤖 [FinEE Llama 8B](https://huggingface.co/Ranjit0034/finee-llama-8b) - Fine-tuned model
+- 📦 [FinEE Package](https://pypi.org/project/finee/) - Python library
+- 💻 [GitHub](https://github.com/Ranjitbehera0034/Finance-Entity-Extractor)