Key points are not available for this paper at this time.
Quality in Wikipedia is enforced through a set of editing policies and guidelines recommended for Wikipedia editors. Neutral point of view (NPOV) is one of the main principles in Wikipedia, which ensures that for controversial information all possible points of view are represented proportionally. Furthermore, language used in Wikipedia should be neutral and not opinionated. However, due to the large number of Wikipedia articles and its operating principle based on a voluntary basis of Wikipedia editors; quality assurances and Wikipedia guidelines cannot always be enforced. Currently, there are more than 40,000 articles, which are flagged with NPOV or similar quality tags. Furthermore, these represent only the portion of articles for which such quality issues are explicitly flagged by the Wikipedia editors, however, the real number may be higher considering that only a small percentage of articles are of good quality or featured as categorized by Wikipedia. In this work, we focus on the case of language bias at the sentence level in Wikipedia. Language bias is a hard problem, as it represents a subjective task and usually the linguistic cues are subtle and can be determined only through its context. We propose a supervised classification approach, which relies on an automatically created lexicon of bias words, and other syntactical and semantic characteristics of biased statements. We experimentally evaluate our approach on a dataset consisting of biased and unbiased statements, and show that we are able to detect biased statements with an accuracy of 74%. Furthermore, we show that competitors that determine bias words are not suitable for detecting biased statements, which we outperform with a relative improvement of over 20%.
Hube et al. (Mon,) studied this question.