Wikipedia is one of the most visited websites in the world. On Wikipedia, Conflict-of-Interest (CoI) editing happens when an editor uses Wikipedia to advance their interests or relationships. This includes paid editing done by organisations for public relations purposes, etc. CoI detection is highly subjective and though closely related to vandalism and bias detection, it is a more difficult problem. In this paper, we frame CoI detection as a binary classification problem and explore various features which can be used to train supervised classifiers for CoI detection on Wikipedia articles. Our experimental results show that the best F-measure achieved is 0.67 by training SVM from a combination of features including stylometric, bias and emotion features. As we are not certain that our non-CoI set does not contain any CoI articles, we have also explored the use of one-class classification for CoI detection. The results show that using stylometric features outperforms other types of features or a combination of them and gives an F-measure of 0.63. Also, while binary classifiers give higher recall values (0.81~0.94), one-class classifier attains higher precision values (0.69~0.74).
@InProceedings{ORIZU18.256, author = {Udochukwu Orizu and Yulan He}, title = "{Content-Based Conflict of Interest Detection on Wikipedia}", booktitle = {Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)}, year = {2018}, month = {May 7-12, 2018}, address = {Miyazaki, Japan}, editor = {Nicoletta Calzolari (Conference chair) and Khalid Choukri and Christopher Cieri and Thierry Declerck and Sara Goggi and Koiti Hasida and Hitoshi Isahara and Bente Maegaard and Joseph Mariani and Hélène Mazo and Asuncion Moreno and Jan Odijk and Stelios Piperidis and Takenobu Tokunaga}, publisher = {European Language Resources Association (ELRA)}, isbn = {979-10-95546-00-9}, language = {english} }
Summary of the paper