Ubuntu

该数据集是从Ubuntu IRC频道提取的大规模现实世界对话生成与检索数据集，包含了两个参与者之间超过三回合的对话。为了减轻由于截断造成的信息丢失，每个对话的上下文被按逆序拼接成一段文本。该数据集规模庞大，其候选发言池包含120千条数据，适用于对话生成和检索任务。

This dataset is a large-scale real-world dialogue generation and retrieval dataset extracted from Ubuntu IRC channels, consisting of conversations with more than three conversational turns between two participants. To mitigate information loss caused by truncation, the context of each dialogue is concatenated into a single text segment in reverse order. This large-scale dataset has a candidate utterance pool containing 120 thousand entries, making it suitable for dialogue generation and retrieval tasks.