Bipartite editing prediction in wikipedia

From WikiPapers
Jump to: navigation, search

Bipartite editing prediction in wikipedia is a 2014 journal article written in English by Chang Y.-J., Tsai Y.-C., Kao H.-Y. and published in Journal of Information Science and Engineering.

[edit] Abstract

Link prediction problems aim to project future interactions among members in a social network that have not communicated with each other in the past. Classical approaches for link prediction usually use local information, which considers the similarity of two nodes, or structural information such as the immediate neighborhood. However, when using a bipartite graph to represent activity, there is no straightforward similarity measurement between two linking nodes. However, when a bipartite graph shows two nodes of different types, they will not have any common neighbors, so the local model will need to be adjusted if the users' goal is to predict bipartite relations. In addition to local information regarding similarity, when dealing with link predictions in a social network, it is natural to employ community information to improve the prediction accuracy. In this paper, we address the link prediction problem in the bipartite editing graph used in Wikipedia and also examine the structure of community in this edit graph. As Wikipedia is one of the successful member-maintained online communities, extracting the community information and solving its bipartite link prediction problem will shed light on the process of content creation. In addition, to the best of our knowledge, the problem of using community information in bipartite for predicting the link occurrence has not been clearly addressed. Hence we have designed and integrated two bipartite-specific approaches to predict the link occurrence: First, the supervised learning approach, which is built around the adjusted features of a local model and, second, the community-awareness approach, which utilizes community information. Experiments conducted on the Wikipedia collection show that in terms of F1-measure, our approaches generates an 11% improvement over the general methods based on the K-Nearest Neighbor. In addition to this, we also investigate the structure of communities in the editing network and suggest a different approach to examining the communities involved in Wikipedia.

[edit] References

This section requires expansion. Please, help!

Cited by

Probably, this publication is cited by others, but there are no articles available for them in WikiPapers.