TweetEmoji-EN

该数据集包含了从推特上收集的英文推文，每条推文至少包含一个表情符号，这些表情符号被用作社会语用学任务的代理标签。数据集分为训练集和验证集，其中训练集包含3100万条推文，验证集包含100万条推文。此外，数据集还包含了三种额外语言（阿拉伯语、意大利语和西班牙语）的数据，并探讨了标签的使用效用。该数据集的规模为3200万条推文，包含1067个独特的表情符号。相关任务包括情感识别、仇恨言论检测、幽默识别、讽刺检测等社会语用学任务。

This dataset contains English tweets collected from Twitter, where each tweet includes at least one emoji, and these emojis serve as proxy labels for sociopragmatic tasks. The dataset is split into training and validation sets, with the training set comprising 31 million tweets and the validation set containing 1 million tweets. Furthermore, the dataset encompasses data in three additional languages: Arabic, Italian, and Spanish, and the utility of these proxy labels is investigated. The total scale of this dataset reaches 32 million tweets and includes 1,067 unique emojis. Relevant tasks include sociopragmatic tasks such as sentiment recognition, hate speech detection, humor identification, and sarcasm detection.