ReINTEL

本文报道了 ReINTEL Shared
社交网站责任信息识别任务，即
在第七届年度研讨会上主办
越南语和语音处理
（VLSP 2020）。给定一条具有各自文本、视觉内容和元数据的新闻，
参与者需要分类是否
新闻是“可靠的”或“不可靠的”。为了
为了产生一个公平的基准，我们引入了一个
超过 10,000 个新的人工注释数据集
从越南社交网络收集的新闻。所有模型都将根据术语进行评估
AUC-ROC 分数，一个典型的分类评估指标。比赛进行了
在 Codalab 平台上。两个月内，
挑战赛吸引了 60 多名参与者
并记录了近 1,000 个提交条目。

This paper reports the ReINTEL Shared Task on Responsible Information Identification on Social Networking Sites, which was hosted at the 7th Annual Workshop on Vietnamese Language and Speech Processing (VLSP 2020). Given a news item with its corresponding text, visual content and metadata, participants are tasked with classifying whether the news is "reliable" or "unreliable". To establish a fair benchmark, we introduce a newly human-annotated dataset consisting of over 10,000 news items collected from Vietnamese social networks. All models will be evaluated using the AUC-ROC score, a standard classification evaluation metric. The shared task was conducted on the Codalab platform. Over a two-month period, the challenge attracted more than 60 participants and received nearly 1,000 submission entries.

搜集汇总

背景与挑战

背景概述

ReINTEL是一个多模态数据集，用于社交网站责任信息识别任务，包含超过10,000条人工注释的越南语新闻，旨在分类新闻的可靠性。该数据集在VLSP 2020研讨会上发布，吸引了广泛参与，评估基于AUC-ROC分数。