匿名化网络公害行为主体刻画数据集

匿名化网络公害行为主体刻画数据集主要面向国家对网络空间公害治理的战略需求，从约200万Telegram用户数据中，基于发送消息数量超过20条、至少加入3个群组、且画像刻画标签大于三种的筛选条件，初步筛选出约6000个用户，然后请3位标注人员依据6000个用户属性信息以及用户对应的发言信息，人工筛选5440个用户数据并修正用户画像构建5个匿名化网络公害行为主体刻画测试数据集，每个数据集包含1024个用户且数据集包括各自的用户属性数据、用户发送消息数据、用户画像刻画基准，数据集之间重复比例平均不超过10%，数据大于200MBThe anonymized dataset for characterizing cyber public nuisance perpetrators is developed to meet the national strategic needs of governing public nuisances in cyberspace. Initially, approximately 6,000 users were screened from a dataset of approximately 2 million Telegram user records, following three filtering criteria: having sent more than 20 messages, joining at least 3 groups, and possessing more than 3 profile characterization tags. Subsequently, 3 annotators manually selected 5,440 valid user records based on the attribute information and corresponding message content of the 6,000 initially screened users, revised the user profiles, and constructed 5 test datasets for characterizing anonymized cyber public nuisance perpetrators. Each test dataset contains 1,024 users, and includes user attribute data, user message sending records, and profile characterization benchmarks. The average inter-dataset duplication rate does not exceed 10%, and the total data size exceeds 200 MB.

搜集汇总

背景与挑战

背景概述

该数据集面向网络空间公害治理需求，从海量Telegram用户数据中筛选并人工标注构建了5个匿名化测试集，每个包含1024个用户及其属性、消息和画像基准。数据量超过200MB，旨在支持网络公害行为主体的刻画研究。