Title :
A New Approach to Detecting Content Anomalies in Wikipedia
Author :
Sinanc, Duygu ; Yavanoglu, Uraz
Author_Institution :
Dept. of Comput., Gazi Univ., Ankara, Turkey
Abstract :
The rapid growth of the web has caused to availability of data effective if its content is well organized. Despite the fact that Wikipedia is the biggest encyclopedia on the web, its quality is suspect due to its Open Editing Schemas (OES). In this study, zoology and botany pages are selected in English Wikipedia and their html contents are converted to text then Artificial Neural Network (ANN) is used for classification to prevent disinformation or misinformation. After the train phase, some irrelevant words added in the content about politics or terrorism in proportion to the size of the text. By the time unsuitable content is added in a page until the moderators´ intervention, the proposed system realized the error via wrong categorization. The results have shown that, when words number 2% of the content is added anomaly rate begins to cross the 50% border.
Keywords :
Internet; Web sites; botany; data mining; hypermedia markup languages; neural nets; pattern classification; text analysis; text editing; zoology; ANN; English Wikipedia; HTML contents; OES; Web mining techniques; anomaly rate; artificial neural network; botany pages; content anomaly detection; encyclopedia; open editing schemas; politics; terrorism; text classification; train phase; wrong categorization; zoology pages; Artificial neural networks; Electronic publishing; Encyclopedias; Internet; Web pages; artificial neural networks; class mapping; data mining; open editing schemas; web classification;
Conference_Titel :
Machine Learning and Applications (ICMLA), 2013 12th International Conference on
Conference_Location :
Miami, FL
DOI :
10.1109/ICMLA.2013.137
