README.md

---
language:
- syr
tags:
- translation
- syriac
- vocalization
- diacritics
- biblical
- eastern-syriac
- mossul
- seq2seq
- marianmt
license: apache-2.0
datasets:
- custom
metrics:
- bleu
- chrf
- accuracy
base_model: Helsinki-NLP/opus-mt-tc-bible-big-sem-en
model-index:
- name: johnlockejrr/marianmt_syr_voc_eastern
  results:
  - task:
      type: text-to-text-generation
      name: Eastern Syriac Vocalization
    dataset:
      name: Eastern Syriac Vocalization Dataset
      type: custom
    metrics:
    - type: bleu
      value: 62.41
      name: BLEU Score
    - type: chrf
      value: 87.98
      name: chrF Score
    - type: accuracy
      value: 58.81
      name: Character Accuracy
---

# MarianMT Eastern Syriac Vocalization Model

A fine-tuned MarianMT model for automatic Eastern Syriac (Mossul Bible) vocalization, converting consonantal (unvocalized) Syriac text to fully vocalized text with diacritical marks.

## Model Description

This model is fine-tuned from [`Helsinki-NLP/opus-mt-tc-bible-big-sem-en`](https://huggingface.co/Helsinki-NLP/opus-mt-tc-bible-big-sem-en) to perform Eastern Syriac vocalization—the task of adding diacritical marks (vowels) to consonantal Syriac text. The model is specifically trained on **Eastern Syriac** texts, and is optimized for the Eastern Syriac vocalization system.

### Key Features

- **Single-direction model**: Converts consonantal Syriac (`>>syr_cons<<`) to vocalized Eastern Syriac (`>>syr_voc<<`)
- **Eastern Syriac optimized**: Trained specifically on Eastern Syriac texts (Mossul edition) and Digital Syriac Corpus texts vocalized in Eastern Syriac
- **High performance**: Achieves 62.41 BLEU, 87.98 chrF, and 58.81% character accuracy on test set
- **Biblical and corpus text optimized**: Trained on Eastern Syriac Bible texts (Mossul edition) and Digital Syriac Corpus texts

## Model Details

### Model Information

- **Architecture**: MarianMT (Transformer-based sequence-to-sequence)
- **Base Model**: `Helsinki-NLP/opus-mt-tc-bible-big-sem-en`
- **Parameters**: 240,944,128 (~241M)
- **Vocabulary Size**: 61,025 tokens
- **Language Tags**: 
  - Source: `>>syr_cons<<` (consonantal Syriac)
  - Target: `>>syr_voc<<` (vocalized Eastern Syriac)

### Training Data

- **Training Examples**: 26,633
- **Validation Examples**: 3,097
- **Test Examples**: 1,239
- **Total**: 30,969 sentence pairs
- **Source**: 
  - Eastern Syriac Bible texts (Mossul edition)
  - Digital Syriac Corpus texts vocalized in Eastern Syriac
- **Format**: Consonantal and vocalized Eastern Syriac pairs

### Training Configuration

- **Batch Size**: 4
- **Effective Batch Size**: 16 (with gradient accumulation)
- **Learning Rate**: 1e-5
- **Max Input/Target Length**: 512 tokens
- **Training Steps**: 66,000 (early stopping)
- **Epochs**: 39.64
- **Optimizer**: AdamW with cosine learning rate schedule
- **Precision**: bfloat16
- **Early Stopping**: Based on validation metrics
- **Training Time**: ~2 days, 13 hours, 51 minutes

### Performance

#### Best Validation Metrics (Epoch 36.04)
- **BLEU**: 63.04
- **chrF**: 88.19
- **Character Accuracy**: 57.84%
- **Validation Loss**: 0.0769

#### Final Test Metrics
- **BLEU**: **62.41**
- **chrF**: **87.98**
- **Character Accuracy**: **58.81%**
- **Test Loss**: 0.0782

## Usage

### Direct Usage

```python
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

tokenizer = AutoTokenizer.from_pretrained("johnlockejrr/marianmt_syr_voc_eastern")
model = AutoModelForSeq2SeqLM.from_pretrained("johnlockejrr/marianmt_syr_voc_eastern")

# Input: consonantal Syriac text
text = "ܒܪܫܝܬ ܐܝܬܘܗܝ ܗܘܐ ܡܠܬܐ"

# Add language tag
input_text = f">>syr_cons<< {text}"

# Tokenize
inputs = tokenizer(input_text, return_tensors="pt", padding=True, truncation=True, max_length=512)

# Generate
outputs = model.generate(**inputs, max_length=512, num_beams=4, length_penalty=0.6)

# Decode
vocalized = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(vocalized)
```

### Using the Pipeline

```python
from transformers import pipeline

vocalizer = pipeline("text2text-generation", 
                     model="johnlockejrr/marianmt_syr_voc_eastern",
                     tokenizer="johnlockejrr/marianmt_syr_voc_eastern")

# Input text (consonantal)
text = "ܒܪܫܝܬ ܐܝܬܘܗܝ ܗܘܐ ܡܠܬܐ"
input_text = f">>syr_cons<< {text}"

# Vocalize
result = vocalizer(input_text, max_length=512, num_beams=4, length_penalty=0.6)
print(result[0]['generated_text'])
```

### Text Normalization

The model expects input text to be normalized to NFC (Normalization Form Composed) Unicode format. The model automatically handles this, but for best results, ensure your input text is properly normalized:

```python
import unicodedata

def normalize_text(text: str) -> str:
    """Normalize text to NFC format."""
    return unicodedata.normalize("NFC", text)

# Normalize input before processing
text = normalize_text("ܒܪܫܝܬ ܐܝܬܘܗܝ")
```

### Input Cleaning

For optimal results, input text should contain only consonantal Syriac characters. The model is designed to work with raw consonantal text, but it can handle text with some punctuation. For best performance, remove vocalization marks from input text if present.

## Generation Parameters

Recommended generation parameters:

- **num_beams**: 4 (beam search for better quality)
- **length_penalty**: 0.6 (encourages longer outputs)
- **early_stopping**: True
- **max_length**: 512 (matches training configuration)
- **do_sample**: False (deterministic generation)

## Limitations and Bias

- **Dialect Specificity**: This model is trained specifically on Eastern Syriac (Mossul edition). Performance may vary on Western Syriac or other Syriac dialects.
- **Domain Specificity**: This model is trained primarily on biblical and corpus Syriac texts. Performance may vary on other domains (e.g., modern Syriac, poetry, prose).
- **Single Direction**: The model only vocalizes consonantal text. It does not perform the reverse operation (removing vocalization).
- **Length Constraints**: Maximum input/output length is 512 tokens. Longer texts should be split into smaller segments.
- **Character Accuracy**: While BLEU and chrF scores are high, character-level accuracy is ~59%, meaning some diacritical marks may be missing or incorrect in complex cases.

## Training Procedure

### Training Infrastructure

- **Hardware**: GPU (CUDA)
- **Training Time**: ~2 days, 13 hours, 51 minutes
- **Framework**: Hugging Face Transformers
- **Evaluation Frequency**: Every 1,000 steps

### Preprocessing

- Text normalized to NFC Unicode format
- Language tags (`>>syr_cons<<` and `>>syr_voc<<`) added to tokenizer vocabulary
- Tokenization using SentencePiece (inherited from base model)

### Hyperparameters

```json
{
  "learning_rate": 1e-5,
  "batch_size": 4,
  "gradient_accumulation_steps": 4,
  "num_epochs": 100,
  "max_input_length": 512,
  "max_target_length": 512,
  "warmup_steps": 1000,
  "weight_decay": 0.01,
  "eval_steps": 1000,
  "save_steps": 1000,
  "save_total_limit": 3
}
```

## Evaluation

The model is evaluated using three metrics:

1. **BLEU Score**: Measures n-gram precision between generated and reference text
2. **chrF Score**: Character-level F-score, more lenient than BLEU
3. **Character Accuracy**: Exact character match percentage

### Evaluation Results

| Metric | Validation (Best) | Test (Final) |
|--------|-------------------|-------------|
| BLEU   | 63.04             | 62.41       |
| chrF   | 88.19             | 87.98       |
| Char Acc | 57.84%          | 58.81%      |
| Loss   | 0.0769            | 0.0782      |

## Citation

If you use this model, please cite:

```bibtex
@misc{marianmt_syr_voc_eastern,
  title={MarianMT Eastern Syriac Vocalization Model},
  author={johnlockejrr},
  year={2025},
  howpublished={\url{https://huggingface.co/johnlockejrr/marianmt_syr_voc_eastern}},
  note={Fine-tuned from Helsinki-NLP/opus-mt-tc-bible-big-sem-en. Trained on Eastern Syriac Bible texts (Mossul) and Digital Syriac Corpus texts.}
}
```

## Acknowledgments

- **Base Model**: [Helsinki-NLP/opus-mt-tc-bible-big-sem-en](https://huggingface.co/Helsinki-NLP/opus-mt-tc-bible-big-sem-en) by the Helsinki NLP team
- **Framework**: [Hugging Face Transformers](https://github.com/huggingface/transformers)
- **Training Framework**: MarianMT architecture
- **Training Data**: Eastern Syriac Bible texts (Mossul edition) and Digital Syriac Corpus texts

## Model Card Contact

For questions, issues, or contributions, please open an issue on the model repository.

## License

This model is released under the Apache 2.0 license, consistent with the base model.