Basma2423/Text-with-Diacritics-Correction

这是一个由Abdou的阿拉伯文字符标注数据集衍生而来的阿拉伯语数据集，经过了预处理和错误注入，用于训练语音和打字更正模型。数据集包含超过150万条文本样本，适用于文本校正、语音识别和阿拉伯语自然语言处理研究。数据集经过了清理，移除了非阿拉伯字符、符号、表情、括号和数字，并引入了四种模拟现实世界中语音和打字错误的错误类型。

This is an Arabic dataset derived from Abdous Arabic Tashkeel Dataset, with additional preprocessing and error injection to assist in training models for speech and typing correction. The dataset contains over 1.5 million text samples and is suitable for text correction, speech recognition, and Arabic NLP research. The dataset has been cleaned, removing non-Arabic characters, symbols, emojis, brackets, and numbers, and introduces four types of errors that simulate real-world speech and typing mistakes.