ROAST

该数据集包含了从五个不同领域精心挑选的6个新数据集，这些数据集用三种语言编写，旨在评估方面基础情感分析中的ROAST任务。此外，该数据集覆盖了多种语言和领域，并包含了详细的标注过程以及标注者之间一致性的度量指标。规模上，它包含了10,304条评论，其中包含63,313个四元组和44,111个类别-极性元组。该数据集的任务是针对评论级别的观点、方面和情感目标检测。

This dataset includes 6 newly curated datasets selected from five distinct domains, written in three languages, and is designed to evaluate the ROAST task in aspect-based sentiment analysis. Additionally, this dataset covers multiple languages and domains, and features detailed annotation procedures as well as inter-annotator agreement metrics. In terms of scale, it contains 10,304 reviews, including 63,313 quadruples and 44,111 category-polarity tuples. The core task of this dataset is to perform review-level opinion, aspect, and sentiment target detection.