SPACE 和 OPOSUM+
收藏资源简介:
SPACE和OPOSUM+是两个用于意见摘要的数据集。SPACE包含酒店领域的评论,而OPOSUM+则包含来自六个不同领域的亚马逊产品评论。这两个数据集都包含大量的原始评论和少量的人工标注的开发和测试集,用于评估。数据集的构建采用了Leave-One-Out (LOO) 方法,通过随机抽样评论元素作为伪摘要,然后从剩余的评论元素中选择输入评论。这些数据集主要用于训练和评估无监督的意见摘要模型,特别是针对特定方面的意见摘要。
SPACE and OPOSUM+ are two datasets designed for opinion summarization. SPACE contains hotel-domain reviews, while OPOSUM+ includes Amazon product reviews across six distinct domains. Both datasets feature a large volume of raw reviews alongside a small number of manually annotated development and test sets for evaluation. The datasets are constructed using the Leave-One-Out (LOO) method: a review element is first randomly sampled as the pseudo-summary, then the input review is selected from the remaining review elements. These datasets are primarily employed for training and evaluating unsupervised opinion summarization models, particularly aspect-based opinion summarization models.




