sk-community/romanized-hindi-transliterator-60M

Model Card for Romanized Hindi Transliterator

Model Summary

This model is a Marian-based Seq2Seq transliterator trained on the Romanized Hindi dataset (1.8M pairs).
It maps Hindi text written in the Roman alphabet (Hindish) back into Devanagari script.

• Architecture: MarianMT (Seq2Seq Transformer)
• Parameters: ~60M
• Training Data: romanized_hindi dataset (IAST + noisy social media + Dakshina)
• Languages: Hindi (Romanized ↔ Devanagari)

Intended Use

• Converting Romanized Hindi to Devanagari for:
  • NLP preprocessing
  • Search & IR
  • Chatbots and voice assistants
  • Sentiment analysis

Example Usage

from transformers import MarianMTModel, MarianTokenizer

model_name = "sk-community/romanized_hindi_transliterator"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)

input_text = "tum kaisa ho"
inputs = tokenizer(input_text, return_tensors="pt")
outputs = model.generate(**inputs)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))
# Output: "तुम कैसा हो"

Performance

• BLEU (Hindi): 77.57
• CER: 0.09
• WER: 0.13

Citation

@article{gharami2025indotranslit,
  title={Modeling Romanized Hindi and Bengali: Dataset Creation and Multilingual LLM Integration},
  author={Kanchon Gharami and Quazi Sarwar Muhtaseem and Deepti Gupta and Lavanya Elluri and Shafika Showkat Moni},
  year={2025}
}