ACM TechNews

How (Much) to Trust Wikipedia

CITRIS Newsletter (12/07) Slack, Gordy

A major issue with the online encyclopedia Wikipedia is assuring the accuracy of its entries, and a team of engineers led by UC Santa Cruz computer scientist Luca de Alfaro has developed a software tool for rating how trustworthy such entries are using the reputation of the contributor as a determining factor. The program is undergirded by a system that assigns and updates a numerical value of reputation to each Wikipedia author, and that value increases as authors make contributions that are preserved by subsequent editors. "The reputation we compute for authors is a good predictor of future behavior: authors with high reputations really do tend to make longer-lasting contributions to Wikipedia," notes de Alfaro. The number of edits that Wikipedia text has been subjected to over time is used to compute the text's trustworthiness, and De Alfaro's system tags each word with a trust value gleaned from the reputation of the word's author, and from the reputation of all visitors who edited nearby text. The text's background is color-coded to reflect the author's reputation value, with clear backgrounds representing the highest value. Edits to text are represented in a shade of orange, with darker shades reflecting lower levels of trustworthiness. "We are essentially automating the usual process of text revision: for each piece of text, we take into account all the people who revised it, giving more weight to people of higher reputation." de Alfaro explains.

http://www.citris-uc.org/December_2007_newsletter#article-5267

---

© Copyright 2008 Information, Inc. This service may be reproduced for internal distribution.