DeepSynth - MLSUM German News Summarization
Dataset Description
Large-scale German news summarization dataset from major German newspapers.
Handles German-specific characters (umlauts, ß) through proper Unicode font rendering.
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… See the full description on the dataset page: https://huggingface.co/datasets/baconnier/deepsynth-de.