VLUS06/VietOnlineNews

VietOnlineNews是一个越南在线新闻数据集，专为单标签多类主题分类任务而构建。每个样本对应一篇新闻文章，并通过`category`字段分配一个主主题标签。数据集从多个越南在线新闻源（包括VietnamNet、Thanh Niên和Tuổi Trẻ）收集，并经过数据清洗管道处理，以去除不完整记录、标签冲突、重复内容、极短文章以及网络爬虫引入的常见噪声。该数据集的主要目标是支持越南自然语言处理（NLP）的研究和实验，特别是越南新闻文章的主题分类。数据集包含268,118个清理后的样本，分为13个主题类别，时间范围从2023年1月1日至2025年12月31日。核心数据包括`title`（标题）、`description`（描述）、`content`（内容）和`category`（类别）四个主要字段。数据集按`category`标签分层划分为训练集（187,682个样本，占70%）、开发集（40,218个样本，占15%）和测试集（40,218个样本，占15%）。VietOnlineNews is a Vietnamese online news dataset constructed for the task of single-label multi-class topic classification. Each sample corresponds to one news article and is assigned exactly one main topic label through the `category` field. The dataset was collected from multiple Vietnamese online news sources (including VietnamNet, Thanh Niên, and Tuổi Trẻ) and processed through a data cleaning pipeline to remove incomplete records, label conflicts, duplicated content, extremely short articles, and common noise introduced during web crawling. The main objective of this dataset is to support research and experimentation in Vietnamese Natural Language Processing (NLP), especially topic classification of Vietnamese news articles. The dataset contains 268,118 cleaned samples, divided into 13 topic categories, with a time range from 01/01/2023 to 31/12/2025. The core data includes four main fields: `title`, `description`, `content`, and `category`. The dataset is split into train (187,682 samples, 70%), dev (40,218 samples, 15%), and test (40,218 samples, 15%) sets using stratified splitting based on the `category` label.