MRS (Multilingual Reply Suggestion)

MRS是一个多语言回复建议数据集，包含十种语言，旨在帮助用户更快地处理邮件和聊天。数据集从公开的Reddit线程中提取消息回复对，并包含机器翻译的示例。MRS可用于比较两种模型家族：检索模型和生成模型。检索模型从固定集合中选择回复，而生成模型则从头开始生成回复。MRS补充了现有的跨语言泛化基准，这些基准主要关注分类和序列标注任务。数据集的应用领域包括提高邮件和聊天应用的自动化回复建议功能，旨在解决多语言环境下回复建议的挑战。

MRS is a multilingual response suggestion dataset covering ten languages, designed to help users process emails and chats more efficiently. Message-response pairs were extracted from public Reddit threads, and machine-translated examples are included in the dataset. MRS can be used to compare two families of models: retrieval-based models and generative models. Retrieval-based models select responses from a fixed corpus, while generative models generate responses from scratch. MRS complements existing cross-lingual generalization benchmarks, which primarily focus on classification and sequence labeling tasks. The application scenarios of this dataset include enhancing automated response suggestion functions for email and chat applications, aiming to address the challenges of response suggestion in multilingual environments.