Hunting for Troll Comments in News Community Forums
Todor Mihaylov Institute for Computational Linguistics Heidelberg University Heidelberg, Germany \AndPreslav Nakov Qatar Computing Research Institute Hamad bin Khalifa University P.O. box 5825, Doha, Qatar This research started in the Sofia University.
Abstract
There are different definitions of what a troll is. Certainly, a troll can be somebody who teases people to make them angry, or somebody who offends people, or somebody who wants to dominate any single discussion, or somebody who tries to manipulate people’s opinion (sometimes for money), etc. The last definition is the one that dominates the public discourse in Bulgaria and Eastern Europe, and this is our focus in this paper.
中文速览
在东欧国家，"网络水军"（troll）常指受雇操纵舆论的付费账号，针对这一现象，研究者收集了保加利亚最大新闻社区论坛 Dnevnik.bg 的近两百万条评论，并通过泄露的"声誉管理合同"锁定了付费水军评论、通过其他用户的举报锁定了"被指认水军"评论，以此构建有标注的数据集。研究团队提取了词袋、词向量聚类、情感词典、元数据（发帖时间与回复位置）等多类特征，训练逻辑回归分类器来区分水军与普通用户的评论。两类水军的识别准确率均达到约 81–82%，其中元数据特征（发帖时间段、是否工作日）对"被指认水军"最为关键，而情感倾向和词汇选择对识别付费水军更有帮助。这项工作不仅验证了"被公开举报为水军"这一定义的有效性，还表明被指认水军与付费水军在行为模式上高度相似，为自动检测舆论操控行为提供了可落地的技术路径。
原文 arXiv:1911.08113;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1911.08113v1