RedTrans-Bench
收藏资源简介:
RedTrans-Bench是由北京航空航天大学、小红书公司等机构共同构建的社交网络服务(SNS)翻译数据集。该数据集包含2858个经过精心挑选的英语-汉语翻译对,涵盖了用户帖子、评论和多媒体字幕等内容。数据集的构建目的是为了评估和训练机器翻译模型在处理SNS领域文化敏感性内容的能力,如幽默定位、表情符号语义和模因适配等。数据集通过严格的人工审核确保了内容的相关性和可靠性,并针对社交网络的特点进行了预处理和优化。
RedTrans-Bench is a social network service (SNS) translation dataset jointly constructed by Beihang University, Xiaohongshu and other institutions. This dataset includes 2858 carefully selected English-Chinese translation pairs, covering content such as user posts, comments and multimedia subtitles. The dataset is designed to evaluate and train the ability of machine translation models to handle culturally sensitive content in the SNS domain, such as humor localization, emoji semantics, meme adaptation and other similar scenarios. It has passed strict manual audits to guarantee the relevance and reliability of the content, and has been preprocessed and optimized in accordance with the characteristics of social networks.

- 1Redefining Machine Translation on Social Network Services with Large Language Models北京航空航天大学, 小红书公司, 北京邮电大学, 南京大学 · 2025年



