CNewSum
收藏资源简介:
CNewSum是一个大规模的中文新闻摘要数据集,由字节跳动创建,包含304,307篇新闻文档及其人工编写的摘要。数据集中的文档较长,摘要具有高度抽象性,旨在推动当前摘要模型在文档级理解与生成方面的发展。CNewSum的独特之处在于其测试集包含摘要的充分性和可推理性标注,这些标注有助于研究人员分析和定位模型性能瓶颈。数据集主要应用于解决中文文本摘要问题,提供了一个坚实的测试平台。
CNewSum is a large-scale Chinese news summarization dataset created by ByteDance, consisting of 304,307 news documents paired with their manually written summaries. The documents in the dataset are relatively lengthy, and the accompanying summaries are highly abstract, aiming to advance the development of current summarization models in document-level understanding and generation. The unique feature of CNewSum is that its test set includes annotations on the sufficiency and inferability of the summaries, which helps researchers analyze and locate the performance bottlenecks of summarization models. This dataset is mainly used to address Chinese text summarization tasks and provides a robust testbed for relevant research.

- 1CNewSum: A Large-scale Chinese News Summarization Dataset with Human-annotated Adequacy and Deducibility Level字节跳动 · 2021年



