Analysis of Controversial Debates in Online Fora - A Showcase Analysis of the CCSVI Discussion in the DMSG Layperson Forum

Fabian Sudau

Abstract

The nature of controversial debates in online fora is often hard to grasp due to the informal discussion style and the sheer number of contributions. Yet, important insights are buried in these openly accessible resources. We want to analyze a showcase of such a debate quantitatively in order to gain a deeper understanding of the underlying dynamics. The showcase stems from the medical field. It is about the controversial hypothesis of Chronic Cerebrospinal Venous Insufficiency (CCSVI) as a cause for Multiple Sclerosis (MS). The debate is observed in a forum provided by the Deutsche Multiple Sklerose Gesellschaft (Engl.: German MS Society) (DMSG) and targeted at laypersons. Our aim is to understand the roles of the forum users and their preferred references to sources of information better. In order to do so, we develop an Information Retrieval algorithm first, that is based on structural forum data, and is able to distinguish posts discussing CCSVI from irrelevant posts. We optimize the parameters of the algorithm by means of an Evolutionary Algorithm. We asses the referenced domains, then classify and visualize them. We identify references to scientific publications. We assign roles to users based on two distinct feature sets: One is the references posted and the other is a carefully selected feature set describing general user behavior. These roles are assigned by means of a kernelized version of the popular K-Means clustering algorithm. We also analyze the presence of homophily and determine the influence of users based on graphs known from the field of Social Network Analysis. Combining the results of these analyses, we can formulate a broad description of user behavior and relationships, community characteristics, and reference influence.

Further Information

Keywords: 
online forum, discussion board, K-Means clustering, information retrieval, evolutionary algorithm, social network analysis, data mining, visualization
Document Type: 
Master's Theses
Month: 
4
Year: 
2013
  • Publication Details:
    Masterarbeit im Studiengang Angewandte Informatik am Institut für Informatik, ZAI-MSC-2013-04, ISSN 1612-6793, Zentrum für Informatik, Georg-August-Universität Göttingen, 2013.

Main menu 2

2011 © Software Engineering for Distributed Systems Group