XL-Sum

XL-Sum 是一个全面而多样的抽象摘要数据集，包含来自 BBC 的 100 万个专业注释文章-摘要对，使用一组精心设计的启发式方法提取。该数据集涵盖从低资源到高资源的 44 种语言，其中许多语言目前没有可用的公共数据集。正如人工和内在评估所表明的那样，XL-Sum 具有高度抽象性、简洁性和高质量。

XL-Sum is a comprehensive and diverse abstractive summarization dataset that contains 1 million professionally annotated article-summary pairs extracted from BBC using a carefully designed set of heuristics. This dataset covers 44 languages ranging from low-resource to high-resource ones, many of which currently lack publicly available datasets. As validated by both human and intrinsic evaluations, XL-Sum exhibits high abstractiveness, conciseness and excellent quality.

搜集汇总

背景与挑战

背景概述

XL-Sum是一个大规模多语言抽象摘要数据集，包含100万BBC文章-摘要对，涵盖44种语言，具有高度抽象性和高质量。该数据集由罗彻斯特大学等机构联合发布，适用于多语言摘要研究。