Research:Ideas/Assessment of article improvement tasks
The topic of flaw detection has been studied previously, see for instance Anderka's PhD thesis One of the issues when doing flaw detection is whether the articles you are analysing actually have the flaw in question. Flaw templates (e.g. "This article needs more sources") are added and removed by humans, thus there might be that an article no longer has the issue in question because someone fixed it, but nobody has yet checked whether the template still applies.
User:SuggestBot is, as far as we're aware, the first deployed Wikipedia tool that aims to do flaw detection, posting specific suggestions for article improvement tasks to its users on English Wikipedia (for an example, see commons:File:SuggestBot article suggestion example - English with tasks.png). Feedback from our users after we launched the version that does flaw detection indicates that our approach might be flawed, it might be that it needs to take the amount of content in the article into consideration.
We are therefore interested in gathering a dataset of articles from English Wikipedia and run it through a process of human assessment to understand to what extent a given set of tasks apply to these articles. This would then allow us to check whether SuggestBot's algorithms match human assessment and if it does not, attempt to find ways to improve it so it does.
We're looking for:
- Three or five fairly experienced Wikipedia users who can help assess the articles.
- Anyone interested in what we're trying to do and wanting to help out.
Ready to create a project page?
- "Analyzing and Predicting Quality Flaws in User-generated Content: The Case of Wikipedia", M. Anderka, 2013