rrtxt/indonesian-spelling-wiki

该数据集是为了训练和评估拼写校正或文本去噪模型而设计的。它包含原始的干净文本和合成干扰文本的对。干扰文本是通过引入常见的拼写错误（如插入或删除字符，以及添加或删除单词之间和内部的空格）而生成的。

This dataset is designed for training and evaluating spell correction or text denoising models. It contains pairs of original clean text and synthetically corrupted text. The corrupted text is generated by introducing common spelling mistakes such as inserting or removing characters, and adding or removing spaces within and between words.