Chinese Financial Announcements

The ChFinAnn dataset is a publicly available financial event dataset that includes five types of financial events: Equity Freeze, Equity Pledge, Equity Repurchase, Equity Overweight, and Equity Underweight. This dataset consists of 32040 samples and 48000 event records, of which approximately 29% of the documents contain more than one record. In addition, 98% of the records have arguments scattered across different sentences, which poses additional challenges for event extraction. This dataset was built to support document level event extraction asks, particularly in the financial field. It is not only large in scale, but also has the characteristics of multi event recording and argument dispersion, which make it a valuable resource for researching document level event extraction.

ChFinAnn数据集是一套公开可用的金融事件数据集，涵盖股权冻结（Equity Freeze）、股权质押（Equity Pledge）、股权回购（Equity Repurchase）、股权增持（Equity Overweight）与股权减持（Equity Underweight）五类金融事件。该数据集包含32040个样本及48000条事件记录，其中约29%的文档包含多条事件记录。此外，98%的事件论元分散于不同句子中，这为事件抽取任务带来了额外挑战。本数据集专为支持文档级事件抽取（document level event extraction）任务而构建，尤其聚焦金融领域场景；其不仅规模庞大，还具备多事件记录与论元分散的特性，因此成为研究文档级事件抽取的宝贵资源。

搜集汇总

背景与挑战

背景概述

Chinese Financial Announcements（ChFinAnn）是一个公开的金融事件数据集，包含股权冻结、股权质押、股权回购、股权增持和股权减持五种事件类型，规模为32040个样本和48000条事件记录。该数据集具有多事件记录（约29%文档包含多个事件）和参数分散（98%记录参数跨句子）的特点，专为支持金融领域的文档级事件提取研究而设计，是一个大规模且具有挑战性的资源。