Identifying relevant YouTube comments to derive socially augmented user models: A semantically enriched machine learning approach

Abstract

Media resources in social Web spaces trigger social interactions, as they consist of motivating means to create and exchange user-generated content. The massive social content could provide rich resources towards deriving social profiles to augment user models and improve adaptation in simulated learning environments. However, potentially valuable social contributions can be buried within highly noisy content that is irrelevant or spam. This paper sketches a research roadmap toward augmenting user models with key user characteristics derived from social content. It then focuses on the first step: identifying relevant content to create data corpus about a specific activity. A novel, semantically enriched machine learning approach to filter out the noisy content from social media is described. An application on public comments in YouTube on job interview videos has been made to evaluate the approach. Evaluation results, which illustrate the ability of the approach to filter noise and identify relevant social media content, are analysed. © 2012 Springer-Verlag.