VASR

该数据集名为VASR，旨在支持视觉情境类比识别任务。它包含大量候选类比，共计50万个，以及一个经过验证的黄金标准数据集，包含3,820个有效的类比。这些数据集通过众包方式进行标注，并且标注结果展现出人类对标签约有80%的一致性。具体规模上，该数据集拥有500,000个候选类比和3,820个验证过的类比，其任务专注于视觉情境的类比识别。

The dataset named VASR is designed to support the visual scenario analogy recognition task. It encompasses 500,000 candidate analogies and a validated gold-standard dataset with 3,820 valid analogies. These datasets were annotated through crowdsourcing, and the annotation results show that human annotators achieve approximately 80% label agreement. Specifically, this dataset includes 500,000 candidate analogies and 3,820 validated analogies, focusing on visual scenario analogy recognition.