HIFI-KPI
收藏资源简介:
HIFI-KPI数据集是由奥尔堡大学计算机科学系和ALIPES ApS合作创建的,包含约181万段文本和约500万个实体,每个实体都链接到iXBRL特定的计算和呈现分类法中的标签。该数据集旨在促进从非结构化财务文本中提取数值关键性能指标(KPI)。数据集的创建基于2017年至2024年间发布的所有10-K和10-Q财务报告,通过解析iXBRL文档来提取文本片段和嵌入式XBRL标签。HIFI-KPI数据集支持多种下游任务,如文本分类、序列标注、结构化信息提取等,可应用于金融领域的问题解答和风险评估。
The HIFI-KPI Dataset was co-created by the Department of Computer Science, Aalborg University and ALIPES ApS. It contains approximately 1.81 million text segments and around 5 million entities, with each entity linked to tags in the iXBRL-specific calculation and presentation taxonomies. This dataset aims to facilitate the extraction of numerical Key Performance Indicators (KPIs) from unstructured financial texts. Constructed using all 10-K and 10-Q financial reports released between 2017 and 2024, the dataset extracts text segments and embedded XBRL tags by parsing iXBRL documents. The HIFI-KPI Dataset supports multiple downstream tasks including text classification, sequence labeling, structured information extraction and others, and can be applied to question answering and risk assessment in the financial domain.




