Predicting Trust among the Authors of the German Wikipedia

Organisatorisches

• Art der Arbeit: Bachelorarbeit
• Interne Betreuer: Dr. Thomas Gottron, René Pickhardt

Beschreibung

The German Wikipedia provides a framework that enables its authors to build a trust network. This network models confidence and trust among the authors, by allowing them to state explicitly which other users they trust. While the initiative is not been taken up by many authors, a small group of approximately 1,000 users is participating in the German Wikipedia trust network. An interesting research question is, if and to which degree the trust relations in this network can be predicted.

The task would be to extract the current trust network from the German Wikipedia, to identify and compute features of individual users and then entire network and to train prediction models on this data. The success of predicted trust relations can be evaluated using metrics like accuracy and by retaining a part of the network data for testing.

In more detail, the work should cover:

• Features extraction from Wikipedia data

• Application of machine learning techniques for link prediction

• Evaluation of prediction models

Requirements

• Programming skills (XML, tools for String processing)

• Willingness to work with machine learning tools.

• Use of a graph database might be an asset