---
license: cc-by-2.0
datasets:
- mbazaNLP/Kinyarwanda_English_parallel_dataset
- mbazaNLP/NMT_Education_parallel_data_en_kin
- mbazaNLP/NMT_Tourism_parallel_data_en_kin
language:
- en
- rw
library_name: transformers
pipeline_tag: translation
---

# Nllb_finetuned_general_en_kin — English ↔ Kinyarwanda (General Purpose)

General-purpose machine translation model for English ↔ Kinyarwanda.
Fine-tuned from [facebook/nllb-200-distilled-1.3B](https://huggingface.co/facebook/nllb-200-distilled-1.3B).

**Fine-tuning code:** [Digital-Umuganda/twb_nllb_finetuning](https://github.com/Digital-Umuganda/twb_nllb_finetuning)

## Usage

```python
from transformers import pipeline

# English → Kinyarwanda
translator = pipeline(
    "translation",
    model="mbazaNLP/Nllb_finetuned_general_en_kin",
    src_lang="eng_Latn",
    tgt_lang="kin_Latn",
    max_length=400,
)
result = translator("Rwanda is a country in East Africa known for its biodiversity.")
print(result[0]["translation_text"])

# Kinyarwanda → English
translator_rev = pipeline(
    "translation",
    model="mbazaNLP/Nllb_finetuned_general_en_kin",
    src_lang="kin_Latn",
    tgt_lang="eng_Latn",
    max_length=400,
)
result = translator_rev("U Rwanda ni igihugu giri mu Afurika yo Hagati.")
print(result[0]["translation_text"])
```

## Intended Use

**Suitable for:**
- General-purpose English ↔ Kinyarwanda translation
- Applications requiring broad language coverage across domains
- Research baseline for NLLB Kinyarwanda translation

**Not intended for:**
- High-stakes translation without human review
- Specialised domains where the education or tourism models may perform better

## Training

Fine-tuned on a general-purpose corpus in a single phase:
- [mbazaNLP/Kinyarwanda_English_parallel_dataset](https://huggingface.co/datasets/mbazaNLP/Kinyarwanda_English_parallel_dataset)
- [mbazaNLP/NMT_Education_parallel_data_en_kin](https://huggingface.co/datasets/mbazaNLP/NMT_Education_parallel_data_en_kin)
- [mbazaNLP/NMT_Tourism_parallel_data_en_kin](https://huggingface.co/datasets/mbazaNLP/NMT_Tourism_parallel_data_en_kin)

Training hardware: A100 40 GB GPU, 2 epochs.

## Evaluation

<!-- TODO: add BLEU/spBLEU/chrF++ scores from evaluation -->

| Lang. Direction | BLEU | spBLEU | chrF++ | TER |
|-----------------|------|--------|--------|-----|
| Eng → Kin       | —    | —      | —      | —   |
| Kin → Eng       | —    | —      | —      | —   |

## Limitations

- General-purpose training may underperform domain-specific models in education or tourism contexts.
- Low-frequency Kinyarwanda vocabulary and tonal nuances may not be handled accurately.
- Outputs should be reviewed for high-stakes applications.
- Maximum reliable input length is approximately 200 tokens.

## Bias and Fairness

Training data spans multiple domains but may not equally represent all registers of Kinyarwanda. Colloquial or dialectal text may translate with lower quality.

## Bias and Fairness

Machine translation models can reflect and amplify biases present in training data. Known limitations include:

- **Domain bias:** Fine-tuned on specific domain data; performance may be lower on out-of-domain text.
- **Cultural bias:** Idiomatic expressions, gender-neutral constructs, and culturally specific references in English may not translate accurately or naturally into Kinyarwanda.
- **Data source bias:** Training data was sourced from specific platforms; text from other sources or registers may yield lower quality translations.
- **Gender:** English gender-neutral pronouns may be rendered with gendered forms in Kinyarwanda based on distributional patterns in training data.

Validate translation quality on domain-representative samples before deployment in high-stakes contexts (legal, medical, government communications).

## Citation

```bibtex
@misc{mbazaNLP2023nllb_finetuned_general,
  author    = {MBAZA-NLP Community},
  title     = {Nllb\_finetuned\_general\_en\_kin: English--Kinyarwanda Machine Translation (General Purpose)},
  year      = {2023},
  url       = {https://huggingface.co/mbazaNLP/Nllb_finetuned_general_en_kin},
  note      = {Hugging Face model repository}
}
```