Wikipedia revision graph extraction based on n-gram cover

Abstract

During the past decade, mass collaboration systems have emerged and thrived on the World-Wide Web, with numerous user contents generated. As one of such systems, Wikipedia allows users to add and edit articles in this encyclopedic knowledge base and piles of revisions have been contributed. Wikipedia maintains a linear record of edit history with timestamp for each article, which includes precious information on how each article has evolved. However, meaningful revision evolution features like branching and revert are implicit and needed to be reconstructed. Also, existence of merges from multiple ancestors indicates that the edit history shall be modeled as a directed acyclic graph. To address these issues, we propose a revision graph extraction method based on n-gram cover that effectively find branching and revert. We evaluate the accuracy of our method by comparing with manually constructed revision graphs. © 2012 Springer-Verlag.