SummFC
收藏资源简介:
SummFC数据集是由巴黎综合理工学院的研究团队创建,旨在解决现有摘要数据集中事实一致性不足的问题。该数据集通过结合最先进的事实一致性模型,筛选出具有改进的事实一致性的摘要样本。数据集包含从CNN/DM、XSUM和XLSUM中筛选出的433293条数据,主要应用于自动摘要系统的开发和评估,以提高摘要的质量和事实准确性。创建过程中,研究团队利用了多种事实一致性模型进行数据筛选,确保数据集的高质量。
The SummFC dataset was developed by a research team from the École Polytechnique to address the critical issue of inadequate factual consistency in existing summarization datasets. To construct this dataset, the team leveraged state-of-the-art factual consistency models to screen summary samples with improved factual consistency. The dataset contains 433,293 curated samples selected from CNN/DM, XSUM and XLSUM. It is mainly applied to the development and evaluation of automatic summarization systems, with the goal of enhancing the quality and factual accuracy of generated summaries. During the dataset creation process, the research team adopted multiple factual consistency models for filtering, ensuring the high quality of the SummFC dataset.




