NorSumm
收藏资源简介:
NorSumm数据集由卑尔根大学和奥斯陆大学的研究团队创建,包含378条挪威新闻文章的人工摘要。数据集旨在为生成式语言模型的抽象摘要能力提供基准测试,每个新闻文章提供三个不同的候选摘要,分别以Bokmål和Nynorsk两种挪威语书面形式呈现。数据集来源于挪威事件抽取数据集EDEN的子集,包含63篇新闻文章,每篇文章由三名具有新闻学背景的挪威语母语者撰写摘要。摘要的创建过程遵循详细的指南,确保摘要自然、多样且符合新闻学标准。该数据集的应用领域主要集中在挪威语生成式语言模型的评估和优化,旨在解决挪威语新闻摘要生成中的挑战。
The NorSumm dataset, created by a research team from the University of Bergen and the University of Oslo, contains 378 manually crafted abstracts of Norwegian news articles. Designed to serve as a benchmark for the abstracting capabilities of generative language models, the dataset provides three distinct candidate summaries for each news article, presented in both Bokmål and Nynorsk written forms of Norwegian. The dataset is derived from a subset of the Norwegian Event Extraction dataset EDEN, encompassing 63 news articles, each summarized by three Norwegian native speakers with a background in journalism. The abstraction process adheres to detailed guidelines to ensure that the summaries are natural, diverse, and meet the standards of journalism. The primary focus of the dataset's application lies in the evaluation and optimization of Norwegian generative language models, aiming to address the challenges in Norwegian news article summarization.

- 1Benchmarking Abstractive Summarisation: A Dataset of Human-authored Summaries of Norwegian News Articles卑尔根大学, 奥斯陆大学 · 2025年



