ImPaKT

ImPaKT数据集由谷歌研究院创建，专注于开放模式知识库构建，包含约2500个来自C4语料库的购物领域文本片段。数据集经过专业标注，涵盖提取的属性、类型、属性摘要以及复合与原子属性之间的一对多关系和蕴含关系。该数据集旨在为跨多个领域的信息提取和知识库构建提供精细调整的语义解析器。The ImPaKT dataset, created by Google Research, focuses on open-schema knowledge base construction. It contains approximately 2,500 shopping-domain text segments sourced from the C4 corpus. The dataset has undergone professional annotation, covering extracted attributes, attribute types, attribute summaries, as well as one-to-many relationships and entailment relations between composite and atomic attributes. This dataset is designed to provide fine-tuned semantic parsers for information extraction and knowledge base construction across diverse domains.