MediFact

该数据集用于评估融合了少样本学习、弱监督（Snorkel）以及基于变压器架构（BART、PEGASUS和SVMs）的混合模型，在生物医学文本处理任务中的应用，这些任务包括跨度识别和摘要生成。该数据集支持包括分类任务的宏-F1、加权-F1等多种评估指标，以及针对摘要任务的ROUGE、BLEU、Meteor和BERTScore。在PerAnsSumm共享任务中，该数据集助力前12支队伍在100支参赛队伍中脱颖而出，其任务涉及视角分类和摘要。

This dataset is used to evaluate hybrid models that combine few-shot learning, weak supervision (Snorkel), and transformer-based architectures (BART, PEGASUS, and SVMs) for biomedical text processing tasks, including span identification and text summarization. It supports multiple evaluation metrics, such as Macro-F1 and Weighted-F1 for classification tasks, as well as ROUGE, BLEU, Meteor, and BERTScore for summarization tasks. In the PerAnsSumm shared task, this dataset helped the top 12 teams stand out among 100 participating teams, where the tasks involved perspective classification and summarization.