A multilingual dataset of 108,478 sentences across 14 languages for emoji-based text augmentation. In each sentence, selected spans (individual words or fixed multi-word expressions) are identified by character offsets and paired with emoji sequences representing their meaning in context. The dataset supports downstream span detection and emoji generation tasks, and was created using a two-stage LLM annotation pipeline with gpt-5.4 for span marking and… See the full description on the dataset page:
https://huggingface.co/datasets/yagizgencer/emojinize-multilingual.
