Date of Publication :7th November 2015

Abstract: Recent years have witnessed the evolution of Web 2.0, which has drastically increased the volume of community-shared textual resources (posts, comments) and media resources (videos, images) over the web. Moreover, today the internet has become an effective communication platform for people. Because of which many cyberbullying content promoters have also been attracted towards online social networks. Online video sharing websites such as YouTube contain large number of videos and users promoting cyberbullying and harassment. Due to immense popularity, anonymity and fewer restrictions for publication, YouTube is misused by some users to promote cyberbullying and online harassment. Our research presents an approach to identify misdemeanor, harassment resulting in cyberbullying by mining the video metadata. We conduct a study on a training dataset obtained by extracting several videos metadata using YouTube API. We formulate the problem of identifying cyberbullying videos as a search problem and present Shark Search algorithm based approach for cyberbullying detection. We present the result using standard information retrieval metrics such as f-measure, precision and recall. The accuracy of the proposed solution on the sample dataset is 83.65%. Our result favors the requirement of several contextual meta-data like, terms present in the title of the videos, description and comments, video subscribers and likes, number of views, YouTube category, length of videos and content focus in cyberbullying detection.

Reference :