Citation Context

...ia to produce a feature value. The high cost of this computational complexity is sometimes mitigated by the ability to build on earlier computations, using incremental calculations.– User reputation =-=[7]-=- (Historic only 10 ) User reputation as computed by WikiTrust [22]. The intuition is that users who have a history of good contributions, and therefore high reputation, are unlikely to commit vandalis...

Citation Context

...rtain characteristic of e that may indicate vandalism. The classifier maps the feature vectors onto {0, 1}, where 0 denotes a regular edit and 1 a vandalism edit. Similarly, some classifiers map onto =-=[0, 1]-=- instead, where values between 0 and 1 denote the classifier’s confidence. To obtain a binary decision a threshold τ ∈ [0, 1] is applied to map confidence values onto {0, 1}. In both cases the mapping...

Citation Context

...ber of deleted words between editors and used this as a measure of controversy. West et. al. and also Adler et. al. have developed vandalism detection methods based on temporal patterns of edits [8], =-=[9]-=-. In both works the main assumption is that offensive edits are reverted much faster than normal edits, therefore by considering the time interval between an arbitrary edit and its subsequent reverts,...

Citation Context

...he deletion template at the top of the page. 3. RELATED WORK There has been extensive work on the subject of detecting and removing vandalism in Wikipedia, mostly using machine learning; for example, =-=[22, 21, 1, 29, 8, 23, 28]-=-, and [2]. At least two of these ([8] and [29]) have been deployed autonomously or semiautonomously on Wikipedia to aid humans in fighting vandalism. None of these solutions address the issue of unenc...

Citation Context

...e following types of edits: • Anonymous edits. It has been shown that the majority of vandalism edits is caused by anonymous editors, whereas the amount of serious anonymous edits is relatively small =-=[1, 16]-=-. Whether an edit has been made by an anonymous or an registered user is stored in the user field of the revision table, which contains either an IP address (anonymous) or an user name (registered). •...

Citation Context

...m and be learned by machines [10]. By combining content analyses with other information about authors and objective measures of edit quality, reputation systems can be developed to identify vandalism =-=[11]-=-. Without these features, spatio-temporal properties of metadata can be sufficient for machine learning algorithms to detect vandalism [12]. However, machine learning algorithms can be improved by usi...

Citation Context

...ia to produce a feature value. The high cost of this computational complexity is sometimes mitigated by the ability to build on earlier computations, using incremental calculations. – User reputation =-=[7]-=- (Historic only10) User reputation as computed by WikiTrust [22]. The intuition is that users who have a history of good contributions, and therefore high reputation, are unlikely to commit vandalism....

Citation Context

...tely be eradicated. Hence several researchers have focused to minimize vandalism through fostering readers trust on content through the analysis of Wikipedia editor’s behavior and the text’s quality (=-=Adler, Alfaro, and Pye, 2010-=-). Wikipedia is designed in such a way that it has fresh information content rather than existing encyclopedias, because the editors at Wikipedia are more active and many in numbers. This largest ency...

Citation Context

...ia to produce a feature value. The high cost of this computational complexity is sometimes mitigated by the ability to build on earlier computations, using incremental calculations.– User reputation =-=[7]-=- (Historic only 10 ) User reputation as computed by WikiTrust [22]. The intuition is that users who have a history of good contributions, and therefore high reputation, are unlikely to commit vandalis...