XED

XED数据集是由赫尔辛基大学创建的多语言细粒度情感数据集，包含人工标注的芬兰语（25k）和英语句子（30k），以及通过标注投影技术为30种其他语言生成的标注。数据集使用Plutchik的核心情感理论进行标注，增加了中性标签，形成多标签多分类数据集。该数据集主要用于情感分析和情绪检测，通过BERT模型和SVM进行评估，显示出与其他类似数据集相当的性能。XED数据集特别适用于低资源语言的研究，为这些语言提供了新的资源。

The XED dataset is a multilingual fine-grained sentiment dataset created by the University of Helsinki. It contains manually annotated Finnish (25k) and English (30k) sentences, along with annotations generated for 30 other languages via annotation projection techniques. All annotations are based on Plutchik's core emotion theory, with an additional neutral label, forming a multi-label multi-classification dataset. This dataset is primarily used for sentiment analysis and emotion detection. Evaluated using BERT models and SVM, it exhibits performance comparable to other similar datasets. The XED dataset is particularly suitable for research on low-resource languages, providing new resources for these languages.

搜集汇总

背景与挑战

背景概述

XED是一个多语言细粒度情感数据集，包含芬兰语和英语的人工标注句子，以及30种其他语言的标注投影数据，适用于情感分析和低资源语言研究。