KPTimes
收藏资源简介:
关键词生成是预测一组传达源文本主要内容的词汇单元的任务。用于生成关键短语的现有数据集仅适用于学术领域,并且包括非专家注释。在本文中,我们介绍了 KPTimes,这是一个大规模的新闻文本数据集,与编辑器策划的关键词配对。通过探索数据集,我们展示了编辑者如何标记文档,以及它们的注释与现有数据集中的注释有何不同。我们还在 KPTimes 上训练和评估最先进的神经关键词生成模型,以深入了解它们在新闻领域的表现。该数据集可在 https://github.com/ygorg/KPTimes 在线获取。
Keyword generation refers to the task of predicting a set of lexical units that convey the core content of a source text. Existing datasets for keyphrase generation are only applicable to the academic domain and involve non-expert annotations. In this paper, we introduce KPTimes, a large-scale news text dataset paired with editor-curated keywords. By exploring this dataset, we demonstrate how editors annotate documents and how their annotations differ from those in existing datasets. We additionally train and evaluate state-of-the-art neural keyword generation models on KPTimes to gain insights into their performance in the news domain. This dataset is publicly available at https://github.com/ygorg/KPTimes.




