Language Homogeneity in the Japanese Wikipedia

Language Homogeneity in the Japanese Wikipedia Karl-Andr´e Skevik Inferno Nettverk A/S, Forskningsparken, Gaustadall´een 21, NO-0349 OSLO, Norway karls@inet.no Abstract. Wikipedia is a potentially very useful source of information, but intuitively it is difficult to have confidence in the quality of an encyclopedia that anyone can modify. One aspect of correctness is writing style, which we examine in a computer based study of the full Japanese Wikipedia. This is possible because Japanese is a language with clearly distinct writing styles using e.g., different verb forms. We find that the writing style of the Japanese Wikipedia is largely consistent with the style guidelines for the project. Exceptions appear to occur primarily in articles with a small num- ber of changes and editors. Keywords: wikipedia, japanese, nlp 1 Introduction Wikipedia1 is a comprehensive Internet based encyclopedia, which is unusual in that the articles are largely written and maintained by volunteers, many of whom are anonymous. The permissive approach used by Wikipedia does however make it natural to question the correctness of the con- tent, and the possibility of achieving a consistent writing style. An encyclopedia is expected to have a formal writing style consistently applied across all articles and in this paper, we examine the writing style of the entire Japanese Wikipedia using automated analysis. This approach is mo- tivated by the presence in Japanese of very distinct writing styles that can serve to signal different degrees of formality and politeness. The difference in language usage is not limited only to word choice but includes different forms for sentence-final elements. By examining all sentences in Wikipedia, we attempt to identify any inconsistencies in writing style. Section 2 of this article lists related work, and Section 3 discusses Japanese writing styles. In Section 4, our analysis procedure is described, with the classification results presented in Section 5. Some of the Wikipedia articles that did not follow the style guidelines are analyzed in Section 6 and our conclusions are presented in Section 7. 2 Related Work Most of the existing research on Wikipedia falls into one of four categories; research that looks at the procedures for avoiding incorrectness and detecting errors in articles, research that attempts to evaluate content correctness, research that looks at completeness, and research that proposes ways in which the article quality can be automatically analyzed. 2.1 Process evaluation The English Wikipedia community has according to Stvilia et al. (2008) developed extensive pro- cesses to achieve article quality. Instead of error prevention, Wikipedia makes use of techniques that allow problems to be fixed quickly when they occur. Copyright 2010 by Karl-Andr´e Skevik 1 http://www.wikipedia.org PACLIC 24 Proceedings 527 Lorenzen (2006) examined the contents of a Wikipedia page used to report potential vandalism over a period of several months and found that a significant amount of resources are spent on addressing this issue. Detecting this in less popular articles can be difficult, but controversial and frequently vandalized pages are likely to be quickly corrected. Vi´egas et al. (2004) find that vandalized pages are …

Content truncated to 500 words