Topic Modeling as an Analysis Tool to Understand the Impact of the Iraq War on the Iraqi Blogosphere.

摘要

Social media provide a democratized platform for expressing one's opinion or viewpoint. The everyday discussions that people have with their families, friends and colleagues became available through blogging services. The emergence of the blogging activity made the classic ethnographic approaches more difficult to deploy. The use of these methods becomes even more problematic when the available data contain a wide variety of languages. This thesis proposes the use of topic modeling as a method for quantitatively analyzing crawled blogs that were created by Iraqi citizens and active over a period of 8 years since the beginning of the Iraq war in 2003. This document presents how data were collected, the way dominant languages were separated in different datasets, the limits of using classification and clustering techniques, the benefits of employing topic modeling, and the evaluation of this technique.