Automatic Cyberbullying Detection: A Semi-supervised Machine Learning Approach

碩士 === 元智大學 === 資訊管理學系 === 107 === Social media is an indispensable online web platform today. With such a platform, people can more freely and quickly spread many creations or content, including Facebook, Twitter, Instagram, LINE, Youtube, Dcard, etc. Although such a platform can communicate, conne...

Full description

Bibliographic Details
Main Authors: Ting-Yu HUANG, 黃亭瑜
Other Authors: Chin-Sheng Yang
Format: Others
Language:zh-TW
Published: 2019
Online Access:http://ndltd.ncl.edu.tw/handle/3wmc24
Description
Summary:碩士 === 元智大學 === 資訊管理學系 === 107 === Social media is an indispensable online web platform today. With such a platform, people can more freely and quickly spread many creations or content, including Facebook, Twitter, Instagram, LINE, Youtube, Dcard, etc. Although such a platform can communicate, connect, and interact instantly, there are also aggressive behaviors, and cyberbullying is one of them. Because social media can be instantly spread without geographical restrictions, and the characteristics of anonymity,Therefore, many users do not need to indicate their true identity, and can quickly and massively disseminate content, including threatening intimidation, verbal abuse, and other bullying behaviors, so they must actively fight against Internet bullying. Our research uses the social media Formspring Q&A article as a text material, using the swear words provided by NoSwearing as a bad word dictionary, and the text, statistics, readability and emotional features most used in past studies as features of our research. However, because supervised learning requires manual data sets as training materials, it will consume a lot of manpower and time costs. Therefore, it is hoped that through the semi-supervised learning technology, the bullying article detection classifier can be learned.