Identifying Communities with Coherent and Opposing Views

Collaboration capabilities are becoming pervasive in online applications. Internet users have access to tools for collective image annotation and editing (flickr, Picasa, Facebook), text content generation (Wikipedia, Google Docs) and even social coding (github). Arguably, social media and social knowledge systems are among the highest impact applications built around collaboration. Following its launch in 2003, Wikipedia has become the most visited not-for-profit free-content encyclopedia, attracting contributers and users who spend 5 minutes a day on average editing and accessing information. Little research has been directed towards understanding the multifaceted community structure of contributers and the topics they collectively promote. We propose a framework for discovery and characterization of collaborative community structure based on raw content-generation analysis. The key steps of our approach include (i) extracting signed pairwise contributor interactions (ii) modeling the editors community as a signed graph and (iii) analyzing the community structure in tandem with the content evolution. Our analysis reveals groupings of common-interest editors that back each other and collectively criticize other communities.