S2M

S2M数据集是由浙江大学软件学院创建，旨在通过转换单轮对话数据集为多轮对话数据集，以提升对话式问答（CQA）模型的性能。该数据集包含19000条对话，通过三个模块（QA对生成器、QA对重组器和问题重写器）处理原始数据，生成更符合对话场景的多轮对话。S2M数据集的应用领域主要集中在自然语言理解中的对话系统，旨在解决单轮数据集与多轮数据集之间的分布差异问题，提高模型在真实对话环境中的表现。

The S2M Dataset was created by the School of Software Engineering, Zhejiang University. Its core goal is to convert single-turn dialogue datasets into multi-turn dialogue datasets, thereby improving the performance of conversational question answering (CQA) models. This dataset contains 19,000 dialogues, which are generated by processing raw data via three modules: the QA pair generator, the QA pair reorganizer, and the question rewriter, to produce multi-turn dialogues that better fit real conversational scenarios. The S2M Dataset is primarily applied in dialogue systems within natural language understanding (NLU), with the objective of addressing the distribution gap between single-turn and multi-turn datasets and enhancing the performance of models in real-world conversational settings.