面向金融领域的主体事件检测

本次数据主要来自金融领域的公开新闻、报道、微博等，样本包含事件样本和无事件样本，为天池赛题【CCKS2023-面向金融领域的主体事件检测】的数据集，该数据集基于句子粒度的上下文进行公司事件检测，事件包含事件类型和主体要素（即公司主体），句中可能存在多个事件（多个公司主体且每个公司都可能存在多个事件类型标签），并且各类型标注样本分布不均匀，部分类型样本量较少，同时数据集中也给出了大量无事件样本。

This dataset is primarily sourced from publicly available financial news, reports, Weibo posts, and other public materials. Its samples include both event-containing samples and event-free samples, and it is the dataset for the Tianchi competition task [CCKS2023 - Financial Domain-oriented Entity Event Detection]. This dataset conducts corporate event detection based on sentence-level context, where each event consists of an event type and its associated entity elements (i.e., corporate entities). A single sentence may contain multiple events involving multiple corporate entities, and each company may be assigned multiple event type labels. Moreover, the distribution of labeled samples across different event types is uneven, with some event types having extremely limited sample sizes, and the dataset also includes a large volume of event-free samples.

搜集汇总

背景与挑战

背景概述

该数据集是面向金融领域的主体事件检测任务，包含来自新闻、报道和微博的文本数据，涵盖103类风险事件。数据特点包括多事件多主体标注、样本分布不均，并提供训练集、验证集和测试集用于模型开发和评估。