README.md

---
language:
- sw
- en
license: apache-2.0
tags:
- translation
- swahili
- english
- opus-mt
- openchs
datasets:
- nllb
- ccaligned
metrics:
- bleu
- chrf
- comet
---

# Swahili-English Translation Model for Child Helpline Services

## Model Description

This model is a fine-tuned version of `Helsinki-NLP/opus-mt-mul-en` for Swahili-to-English translation, specifically optimized for child helpline call transcriptions in East Africa.

**Developed by:** BITZ IT Consulting Ltd  
**Project:** OpenCHS (Open Child Helpline System)  
**Funded by:** UNICEF Venture Fund  
**License:** Apache 2.0

## Performance

### Test Set (General Translation)
- **BLEU:** 0.2272
- **chrF:** 42.25
- **Improvement over baseline:** +0.0%

### Domain Evaluation (Call Transcriptions)
- **Domain BLEU:** 0.0000
- **Domain chrF:** 2.90
- **Domain COMET-QE:** 0.0000

## Intended Use

**Primary Use Case:** Translating Swahili helpline call transcriptions to English for case documentation, quality assurance, and cross-border referrals.

**Languages:** Swahili (source) → English (target)

## Usage

```python
from transformers import MarianTokenizer, MarianMTModel

model_name = "brendaogutu/sw-en-translation-v1"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)

swahili_text = "Habari za asubuhi. Ninaitwa Amina na nina miaka 14."
inputs = tokenizer(swahili_text, return_tensors="pt", padding=True)
outputs = model.generate(**inputs, num_beams=5, max_length=256)
translation = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(translation)
```

## Training Details

**Base Model:** Helsinki-NLP/opus-mt-mul-en  
**Training Epochs:** 8  
**Batch Size:** 128  
**Learning Rate:** 3e-05  
**Hardware:** NVIDIA GPU with FP16 mixed precision

---

*This model is part of the OpenCHS project supporting child helpline services across East Africa.*