Predicting Wikipedia Editor's Editing Interest Based on Factor Graph Model

Abstract

Recruiting or recommending appropriate latent editors who can edit a specific entry (or called article) plays an important role in improving the quality of Wikipedia entries. To predict an editor's editing interest for Wikipedia entries, this paper proposes an Interest Prediction Factor Graph (IPFG) model, which is characterized by editor's social properties, hyperlinks between Wikipedia entries, categories of an entry and other important features. Furthermore, the paper suggests a parameter learning algorithm based on the gradient descent and Loopy Sum-Product algorithms for factor graphs. The experiment on a Wikipedia dataset shows that, the average prediction accuracy (F1-Measure) of the IPFG model could be up to 87.5%, which is about 35% higher than that of a collaborative filtering approach. Moreover, the paper analyses how incomplete social properties and editing bursts affect the prediction accuracy of the IPFG model. What we found would provide a useful insight into effective Wikipedia article tossing, and improve the quality of those entries that belong to specific categories by means of collective collaboration.