TRACSUM
收藏资源简介:
TRACSUM是一个新的基准数据集,用于医学领域的面向方面的摘要,其中生成的摘要与句子级别的引用配对,使用户能够追溯回原始上下文。该数据集包含500个医学摘要,针对七个关键医学方面进行了标注,产生了3.5K摘要-引用对。数据集创建过程中,首先使用Mistral Large模型自动生成摘要和引用,然后由人类专家进行评估和修订。TRACSUM旨在解决医学文献摘要中的事实准确性问题,通过提供引用源文本,用户可以更容易地定位相关上下文并验证生成的内容。
TRACSUM is a novel benchmark dataset for aspect-based summarization in the medical domain, where generated summaries are paired with sentence-level citations to enable users to trace back to the original context. This dataset contains 500 medical summaries annotated against seven key medical aspects, resulting in 3.5k summary-citation pairs. During its construction, summaries and citations were first automatically generated using the Mistral Large model, then evaluated and revised by human experts. TRACSUM aims to address the issue of factual accuracy in medical literature summarization: by providing the source texts of the citations, users can more easily locate relevant contexts and verify the generated content.

- 1通过德国杜伊斯堡-埃森大学 · 2025年



