ru-kbd-opus

Russian-Kabardian Translation Model (ru-kbd-opus)

Fine-tuned MarianMT model for Russian to Kabardian (East Circassian) translation.

• Developed by: Eduard Emkuzhev
• Model type: Neural Machine Translation (Marian Transformer)
• Language pair: Russian (ru) → Kabardian (kbd)
• License: CC BY-NC 4.0
• Base model: Helsinki-NLP/opus-mt-ru-uk (Apache 2.0)
• Training data: adiga-ai/circassian-parallel-corpus (CC BY 4.0)

Model Description

This model translates from Russian to Kabardian, an endangered Northwest Caucasian language with approximately 500,000 speakers. Kabardian features complex polysynthetic morphology with 50+ consonants and ergative-absolutive alignment.

Intended Use

Primary uses:

• Language preservation and revitalization
• Educational materials translation
• Research in low-resource machine translation
• Cultural heritage digitization
• Supporting Kabardian language learners

Limitations:

• Non-commercial use only (CC BY-NC 4.0)
• Best performance on everyday language
• May struggle with highly technical or literary texts
• Requires proper handling of Kabardian-specific character Ӏ (palochka)

Training Data

Dataset: adiga-ai/circassian-parallel-corpus

• Subset: ru_kbd (Russian → Kabardian)
• Total training examples: ~100K parallel sentence pairs
• Dataset license: CC BY 4.0
• Dataset author: Anzor Qunash (adiga.ai)
• Content: Dictionary entries, folklore, proverbs, everyday expressions

Training Procedure

Base Model

• Architecture: Marian Transformer (transformer-align)
• Base: Helsinki-NLP/opus-mt-ru-uk (Russian-Ukrainian translation)
• Transfer learning: Adapted from Russian-Ukrainian to Russian-Kabardian

Hyperparameters

base_model: Helsinki-NLP/opus-mt-ru-uk
training_examples: 200,000
epochs: 7
batch_size: 32
learning_rate: 3e-4
optimizer: AdamW
max_sequence_length: 128
warmup_steps: 500
weight_decay: 0.01
framework: transformers 4.36.0

Special Preprocessing

The model uses a special character mapping for training:

• Kabardian Ӏ (palochka) → I (Latin I) during training
• I → Ӏ restored during inference

This ensures better tokenization compatibility with the MarianMT tokenizer.

Performance

Benchmark Results

Tested on 1,000 examples from adiga-ai/circassian-parallel-corpus:

Test Configuration:

• Test size: 1,000 examples
• Sampling: Every 50th sentence from corpus
• Generation: beam_search (num_beams=4)
• Device: Apple M-series (MPS)
• Seed: 42 (reproducible)

Translation Examples

Note: The model handles morphologically complex Kabardian words and preserves the special palochka character (Ӏ).

How to Use

Installation

pip install transformers torch sentencepiece

Basic Usage

from transformers import MarianMTModel, MarianTokenizer

# Load model and tokenizer
model_name = "kubataba/ru-kbd-opus"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)

# Translation function
def translate_ru_to_kbd(text):
    inputs = tokenizer(text, return_tensors="pt", padding=True)
    outputs = model.generate(**inputs, max_length=128, num_beams=4)
    translation = tokenizer.decode(outputs[0], skip_special_tokens=True)
    # Restore Kabardian palochka
    return translation.replace('I', 'Ӏ')

# Example
russian_text = "Здравствуйте!"
kabardian_text = translate_ru_to_kbd(russian_text)
print(f"RU: {russian_text}")
print(f"KBD: {kabardian_text}")

Batch Translation

texts = [
    "Добрый день!",
    "Как вас зовут?",
    "Спасибо за помощь"
]

inputs = tokenizer(texts, return_tensors="pt", padding=True, truncation=True)
outputs = model.generate(**inputs, max_length=128, num_beams=4)
translations = [
    tokenizer.decode(out, skip_special_tokens=True).replace('I', 'Ӏ') 
    for out in outputs
]

for src, tgt in zip(texts, translations):
    print(f"RU: {src} → KBD: {tgt}")

Limitations and Bias

• Domain specificity: Best performance on conversational and dictionary-style text
• Technical vocabulary: May struggle with modern technical terms not in training data
• Literary style: Not optimized for highly stylized or poetic translations
• Character encoding: Requires proper UTF-8 support for Kabardian Cyrillic + palochka (Ӏ)
• Low-resource challenges: Limited by available parallel data (~189K pairs)

Ethical Considerations

This model supports preservation of Kabardian, an endangered language classified by UNESCO.

Responsible use:

• Verify translations for critical applications with native speakers
• Use as a tool to support, not replace, human translators
• Respect cultural context when translating culturally significant content
• Acknowledge the model's limitations in formal or official contexts

About Kabardian Language

Kabardian (Adyghe-Kabardian, East Circassian) is a Northwest Caucasian language spoken by approximately 500,000 people, primarily in:

• Kabardino-Balkaria (Russia)
• Karachay-Cherkessia (Russia)
• Turkey (diaspora)
• Middle East (diaspora)

Linguistic features:

• 50+ consonant phonemes (one of the world's most complex phonological systems)
• Polysynthetic morphology
• Ergative-absolutive alignment
• Unique character: Ӏ (palochka) - represents the glottal stop

License and Attribution

This Model

• License: Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0)
• Author: Eduard Emkuzhev
• Year: 2025

Base Model

• Model: Helsinki-NLP/opus-mt-ru-uk
• License: Apache 2.0
• Authors: Language Technology Research Group at the University of Helsinki
• Link: https://huggingface.co/Helsinki-NLP/opus-mt-ru-uk

Training Dataset

• Dataset: Circassian-Russian Parallel Corpus v1.0
• License: CC BY 4.0
• Author: Anzor Qunash (adiga.ai)
• Link: https://huggingface.co/datasets/adiga-ai/circassian-parallel-corpus

Citation

If you use this model in your research, please cite:

@misc{emkuzhev2025rukbd,
  author = {Eduard Emkuzhev},
  title = {Russian-Kabardian Neural Machine Translation Model},
  year = {2025},
  publisher = {HuggingFace},
  howpublished = {\url{https://huggingface.co/kubataba/ru-kbd-opus}}
}

Please also cite the base model and dataset:

@misc{helsinki-nlp-opus-ru-uk,
  author = {Language Technology Research Group at the University of Helsinki},
  title = {OPUS-MT Russian-Ukrainian Translation Model},
  year = {2020},
  publisher = {HuggingFace},
  howpublished = {\url{https://huggingface.co/Helsinki-NLP/opus-mt-ru-uk}}
}

@dataset{qunash2025circassian,
  author = {Anzor Qunash},
  title = {Circassian-Russian Parallel Text Corpus v1.0},
  year = {2025},
  publisher = {adiga.ai},
  url = {https://huggingface.co/datasets/adiga-ai/circassian-parallel-corpus}
}

Acknowledgments

• Helsinki-NLP team for the excellent OPUS-MT base models
• Anzor Qunash (adiga.ai) for creating and publishing the Circassian-Russian Parallel Corpus
• Kabardian language community for preserving and promoting their language
• All contributors to Circassian language digitization efforts

Related Models

• Reverse direction: kubataba/kbd-ru-opus - Kabardian to Russian
• Base model: Helsinki-NLP/opus-mt-ru-uk
• Dataset: adiga-ai/circassian-parallel-corpus

Technical Details

• Framework: PyTorch + Transformers
• Model size: ~300MB
• Vocabulary size: ~57K tokens
• Parameters: ~74M
• Inference: CPU and GPU compatible
• Optimal device: GPU or Apple Silicon (MPS)

Contact

• Author: Eduard Emkuzhev
• Email: [email protected]
• GitHub: https://github.com/kubataba
• Issues: Report issues

For commercial licensing inquiries, please contact via email.

---

Model Card Authors: Eduard Emkuzhev

Last Updated: December 2025

Version: 1.0.1