M-MELD

M-MELD是首个针对英语以外语言的情感识别对话数据集，由马里兰大学创建，包含4504个对话和45871条人类标注的文本对话，涵盖希腊语、波兰语、法语和西班牙语。数据集通过人工翻译确保对话的连贯性和情感表达的准确性。创建过程中，首先使用人类翻译的对话，然后通过专业标注人员匹配原始对话和翻译版本。M-MELD不仅适用于情感识别研究，还可用于多语言情感分类和作为人类标注的平行语料库，用于训练机器翻译系统，旨在解决多语言情感识别的挑战。

M-MELD is the first conversational emotion recognition dataset for languages other than English, developed by the University of Maryland. It consists of 4504 conversations and 45871 manually annotated textual dialogues, covering Greek, Polish, French and Spanish. The dataset ensures the coherence of dialogues and the accuracy of emotional expression through human translation. During the creation process, human-translated conversations were first utilized, then professional annotators matched the original dialogues with their translated versions. M-MELD can be applied not only to emotion recognition research, but also to multilingual emotion classification, and can serve as a manually annotated parallel corpus for training machine translation systems, aiming to address the challenges in multilingual emotion recognition.