Detecting Rumors Transformed from Hong Kong Copypasta

Abstract

A copypasta is a piece of text that is copied and pasted in online forums and social networking sites (SNSs) repeatedly, usually for a humorous or mocking purpose. In recent years, copypasta is also used to spread rumors and false information, which damages not only the reputation of individuals or organizations but also misleads many netizens. This paper presents a tool for Hong Kong netizens to detect text messages that are copypasta or their variants (by transforming an existing copypasta with new subjects and events). We exploit the Encyclopedia of Virtual Communities in Hong Kong (EVCHK), which contains a database of 315 commonly occurred copypasta in Hong Kong, and a CNN model to determine whether a text message is a copypasta or its variant with an accuracy rate of around 98%. We also showed a prototype of a Google Chrome browser extension that provides a user-friendly interface for netizens to identify copypasta and their variants on a selected text message directly (e.g., in an online forum or SNS). This tool can show the source of the corresponding copypasta and highlight their differences (if it is a variant). From a survey, users agreed that our tool can effectively help them to identify copypasta and hence help stop the spreading of this kind of online rumor.