baconnier/deepsynth-fr

DeepSynth - MLSUM French News Summarization

Dataset Description

Large-scale French news summarization dataset from major French newspapers. Enables training multilingual DeepSeek-OCR models with proper Unicode/diacritics handling.

This dataset is part of the DeepSynth project, which uses visual text encoding for multilingual summarization with the DeepSeek-OCR vision-language model. Text documents are converted into images and processed through a frozen 380M parameter visual encoder, enabling 20x token compression while preserving document layout and structure.

Key Features

• —Original High-Quality Images: Full-resolution images stored once, augmented on-the-fly during training
• —Random Augmentation Pipeline: Rotation, perspective, color jitter, and resize transforms for better generalization
• —Visual Text Encoding: 20x compression ratio (1 visual token ≈ 20 text tokens)
• —Document Structure Preservation: Layout and formatting maintained through image representation
• —Human-Written Summaries: High-quality reference summaries for each document
• —Deduplication Tracking: Source dataset and index tracking prevents duplicates

Dataset Statistics

• —Total Samples: ~392,000
• —Language(s): French
• —Domain: French news articles
• —Average Document Length: ~700 tokens
• —Average Summary Length: ~50 tokens

Source Dataset

Based on MLSUM (MultiLingual SUMmarization) French subset.

• —Original Authors: Scialom et al. (2020)
• —Paper: MLSUM: The Multilingual Summarization Corpus
• —License: CC BY-NC-SA 4.0

Image Augmentation Pipeline

Images are stored at original resolution (up to 1600×2200) and augmented during training for better generalization:

Available Augmentation Transforms

• —Random Rotation: ±10° rotation for orientation invariance
• —Random Perspective: 0.1-0.2 distortion to simulate viewing angles
• —Random Resize: 512-1600px range for multi-scale learning
• —Color Jitter: Brightness, contrast, saturation adjustments (±20%)
• —Random Horizontal Flip: Optional (use with caution for text)

All transforms preserve aspect ratio with padding to maintain text readability. This approach:

• —Reduces storage: 6x less disk space (single image vs 6 resolutions)
• —Increases flexibility: Any resolution on-the-fly vs pre-computed fixed sizes
• —Improves generalization: Random transforms prevent overfitting to specific resolutions

Dataset Structure

Data Fields

• —text (string): Original document text
• —summary (string): Human-written summary
• —image (PIL.Image): Original full-size rendered document image (up to 1600×2200)
• —source_dataset (string): Origin dataset name
• —original_split (string): Source split (train/validation/test)
• —original_index (int): Original sample index for deduplication

Data Example

{
    'text': 'Le gouvernement français a annoncé de nouvelles mesures...',
    'summary': 'Nouvelles mesures gouvernementales contre le changement climatique.',
    'image': <PIL.Image>,  # Original resolution (up to 1600×2200)
    'source_dataset': 'MLSUM (fr)',
    'original_split': 'train',
    'original_index': 0
}

Usage

Loading the Dataset

from datasets import load_dataset

# Load full dataset
dataset = load_dataset("baconnier/deepsynth-fr")

# Streaming for large datasets
dataset = load_dataset("baconnier/deepsynth-fr", streaming=True)

Training Example with DeepSeek-OCR and Augmentation

from transformers import AutoProcessor, AutoModelForVision2Seq
from datasets import load_dataset
from deepsynth.data.transforms import create_training_transform

# Load model and processor
model = AutoModelForVision2Seq.from_pretrained("deepseek-ai/DeepSeek-OCR")
processor = AutoProcessor.from_pretrained("deepseek-ai/DeepSeek-OCR")

# Load dataset
dataset = load_dataset("baconnier/deepsynth-fr")

# Create augmentation pipeline (random rotation, perspective, resize, color jitter)
transform = create_training_transform(
    target_size_range=(512, 1600),  # Random resize range
    rotation_degrees=10,             # ±10° rotation
    perspective_distortion=0.1,      # Perspective transform
    brightness_factor=0.2,           # ±20% brightness
    contrast_factor=0.2,             # ±20% contrast
)

# Process sample with augmentation
sample = dataset['train'][0]
augmented_image = transform(sample['image'])  # Apply random transforms
inputs = processor(
    images=augmented_image,
    text=sample['text'],
    return_tensors="pt"
)

# Fine-tune decoder only (freeze encoder)
for param in model.encoder.parameters():
    param.requires_grad = False

# Training loop with on-the-fly augmentation...

Training Recommendations

DeepSeek-OCR Fine-Tuning

# Recommended hyperparameters with augmentation
training_args = {
    "learning_rate": 2e-5,
    "batch_size": 4,
    "gradient_accumulation_steps": 4,
    "num_epochs": 3,
    "mixed_precision": "bf16",
    "freeze_encoder": True,  # IMPORTANT: Only fine-tune decoder

    # Augmentation parameters
    "rotation_degrees": 10,           # Random rotation ±10°
    "perspective_distortion": 0.1,    # Perspective transform
    "resize_range": (512, 1600),      # Random resize 512-1600px
    "brightness_factor": 0.2,         # ±20% brightness
    "contrast_factor": 0.2,           # ±20% contrast
}

Expected Performance

• —Baseline (text-to-text): ROUGE-1 ~40-42
• —DeepSeek-OCR (visual): ROUGE-1 ~44-47 (typical SOTA)
• —Training Time: ~6-8 hours on A100 (80GB) for full dataset
• —GPU Memory: ~40GB with batchsize=4, mixedprecision=bf16

Dataset Creation

This dataset was created using the DeepSynth pipeline:

1. 1.Source Loading: Original text documents from MLSUM (fr)
2. 2.Text-to-Image Conversion: Documents rendered as PNG images (DejaVu Sans 12pt, Unicode support)
3. 3.Original Resolution Storage: Full-quality images stored once (up to 1600×2200)
4. 4.Incremental Upload: Batches of 5,000 samples uploaded to HuggingFace Hub
5. 5.Deduplication: Source tracking prevents duplicate samples

Note: Images are augmented on-the-fly during training using random transformations (rotation, perspective, resize, color jitter) for better generalization across different resolutions and conditions.

Rendering Specifications

• —Font: DejaVu Sans 12pt (full Unicode support for multilingual text)
• —Line Wrapping: 100 characters per line
• —Margin: 40px
• —Background: White (255, 255, 255)
• —Text Color: Black (0, 0, 0)
• —Format: PNG with lossless compression

Citation

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

@misc{deepsynth-fr,
    title={{DeepSynth MLSUM French News Summarization: Visual Text Encoding with Random Augmentation for Summarization}},
    author={Baconnier},
    year={2025},
    publisher={HuggingFace},
    howpublished={\url{https://huggingface.co/datasets/baconnier/deepsynth-fr}}
}

Source Dataset Citation

@inproceedings{scialom2020mlsum,
    title={MLSUM: The Multilingual Summarization Corpus},
    author={Scialom, Thomas and Dray, Paul-Alexis and Lamprier, Sylvain and Piwowarski, Benjamin and Staiano, Jacopo},
    booktitle={Proceedings of EMNLP},
    year={2020}
}

License

Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International

Note: This dataset inherits the license from the original source dataset. Please review the source license before commercial use.

Limitations and Bias

• —French-specific: Requires French language models
• —Diacritics: Proper handling of accents (é, è, ê, etc.) critical
• —Regional: May contain France-specific cultural references
• —News domain: Limited to journalistic style

Additional Information

Dataset Curators

Created by the DeepSynth team as part of multilingual visual summarization research.

Contact

• —Repository: DeepSynth GitHub
• —Issues: GitHub Issues

Acknowledgments

• —DeepSeek-OCR: Visual encoder from DeepSeek AI
• —Source Dataset: MLSUM (fr)
• —HuggingFace: Dataset hosting and infrastructure