Ubuntu Chat Log Conversations

该数据集来源于Freenode IRC网络中与Ubuntu相关的话题特定聊天室，包含了多轮次、双边对话以及特定领域的对话数据。在验证集和测试集中，该数据集的正向三元组与负向三元组的比例达到了1:9，这使得我们可以使用如Recall@k等信息检索指标进行评估。数据规模方面，包含了一百万个训练三元组，五十万个验证三元组和五十万个测试三元组。该数据集的任务是针对多轮对话中的响应选择。

This dataset is sourced from topic-specific chatrooms related to Ubuntu on the Freenode IRC network, containing multi-turn, bilateral domain-specific dialogue data. For the validation and test sets, the ratio of positive to negative triples is 1:9, which enables evaluation using information retrieval metrics such as Recall@k. In terms of data scale, the dataset includes 1,000,000 training triples, 500,000 validation triples, and 500,000 test triples. The task of this dataset is response selection in multi-turn dialogues.