Wikipedia:Researching Wikipedia

Researching Wikipedia (formerly known as State of Wikipedia) discusses some ways to quantitatively measure various aspects of Wikipedia project as well as covers research done in that area. The subject is difficult, as there are different goals that Wikipedia may have, and different ways of measuring achievement of those goals.

Theory

edit

Raw numbers

edit

A hard way of measuring success is to count the number of articles in Wikipedia. This information can be found on the Statistics page. A problem with just counting the number of articles is, what is an "article"? A large percentage of our "articles" may be extremely short stubs, or even just consist of uncaught vandalism. {{merge}}ing stubby articles leads to fewer, better articles, without losing any content. A more accurate measure of the size of Wikipedia is the number of characters or words in articles. Wikipedia as of October 2006 had 1.4 million articles with an average length of 3,300 characters.

Such a measurement gives no indication of the quality of content. It is much more difficult to estimate the number of good, useful, accurate, or balanced articles in Wikipedia. For this, we may only take into account articles that have been in some way assessed, either as "featured", "good", "A-" or "B-Class" articles. As of February 2007, one in ca. 550 articles on Wikipedia is either "featured" or "good".

One way to think about the Statistics page is to consider it a measure of Wikipedia's success as a project rather than as a reference work. Since it is a project for producing a reference work (with community building being a side effect, not a secondary goal), assessment of success of the project will be directly tied to assessment of the reference work.

Relevance to the Web

edit

Another way to consider Wikipedia success is to ask how relevant Wikipedia's information is to the World Wide Web. How many hits per day does the Wikipedia site receive? How many readers come from Google? Which pages have high Google PageRank?

A measure of Wikipedia's popularity is provided by its entry on Alexa which shows its web traffic rankings.

One measure that's valuable, but difficult to automate, is to consider Wikipedia:Top 10 Google hits. Of the subjects already in Wikipedia, how many are good enough references that they rank high on Google?

Yet another measurement might involve the number of, or degree to which, other sites use Wikipedia's content. The fact that a number of other sites trust the accuracy of Wikipedia's content is a strong indicator of its success.

Coverage

edit

Another axis to consider is coverage by Wikipedia. Coverage is a measure of how much of the information we need in Wikipedia is already there. How well does Wikipedia "cover" the range of knowledge that it should?

One way to think of coverage is to imagine some kind of "endpoint" in the future – Edit Zero – where all the information that's Wikipedia-worthy is in the system. At that point, the work of Wikipedians will change from writing about existing subjects, to adding articles about new subjects as new people, events, countries, awards ceremonies, species, albums, books, and planets come into being. A measure of Wikipedia's current coverage would be to measure how many of the articles in that imagined encyclopedia already exist in some useful form.

This is, in most ways, an immeasurable metric. We don't know how many articles will be in Wikipedia at Edit Zero, so we can't know what percentage of those we already have. The best we can hope to do is approximate the "real" coverage metric with some ad hoc measurements.

Some proposed approximations:

  • Of the entries in the 1911 Encyclopædia Britannica, how many have corresponding Wikipedia articles? (Totally crude, but if we were back in 1911, wouldn't we want to have at least as much knowledge as the EB? Close to it?)
  • What percentage of Wikipedia searches come up empty? (This would measure what percentage of things Wikipedia readers think should be in the system are already there.)
  • Of the internal links inside Wikipedia, what percentage point nowhere? How many have non-stub articles at the endpoint? (This would measure what percentage of things Wikipedia authors think should be in the system are already there.)

Note that the Edit Zero model is simplistic in expecting the number of Wikipedia-worthy articles to converge at some point in the future.

List of conducted studies and other resources

edit

Wikipedia (primarily) and other Wikimedia projects are increasingly generating research concerned with studying phenomena responsible for their functioning. Some of that research has been published in professional academic journals, or presented at conferences: see Wikipedia:Academic studies of Wikipedia.

However a significant number of other inquiries are not published in such journals. As a result, some essays in the Wikipedia namespace on Wikipedia, certain pages on our Meta wiki, and similar pages on sister projects have been increasingly filled with such short research papers, essays and other resources. meta:Research is the place where such research is supposed to be coordinated, but in fact the majority of tools and papers can be found on English Wikipedia. Below is a guide to those resources.

Note 1: Most interesting and more-or-less up-to-date projects are bolded.

Note 2: Graphs, charts and such should be added to Category:Wikipedia charts

Keywords:

  • Editors: about editors
  • Users: about users
  • Articles: about articles
  • Technical: technical aspects of the projects (software, code...)
Item Description and comments Last updated as of Time series from Analysis Has tables/lists Has charts Keywords
Administrator statistics: User:NoSeptember/The NoSeptember Admin Project Lots of admin-related stats, many subpages. Feb 2007 To the beginning, as much as possible Yes Yes Yes Editors
Announcements Announcements about 1) Important milestones, statistics and Alexa ranking news concerning the English Wikipedia (see Special:Statistics) 2) Any news concerning the Wikimedia Foundation that affects the English Wikipedia. Monthly Since January 2001 Press release type Yes No
Articles for deletion stats Rough stats about AfD, see subpages. Particularly useful: Wikipedia:AFD 100 days: A computer script designed by Dragons flight was used to parse 100 days of AFD logs from June 1, 2005 – September 8, 2005 searching for bolded keywords (e.g. delete, keep, merge, redirect, kill, cleanup, etc) in signed comments. This has allowed a large statistical sample to be generated from which important patterns in voting and article deletion behavior might be identified. January 2006 February 7 Some Yes No Articles
Articles per population The number of Wikipedia articles that exist in a language per million total speakers of that language September 2006 None but history shows series of updates from November 2005 No Yes No Articles
Awareness statistics Attempts to measure the growth in public awareness of Wikipedia. Primarily concerned with Wikipedia's Alexa ratings, i.e. 'how popular Wikipedia is'. Varies, but most tables are up to date as of 7 January 2007; charts are labelled January 2006 Varies, from October 2002 and later Yes Yes Yes Users
Browsers What browsers are used to access English Wikipedia September 2004 History shows old data from April 2004 No Yes No Technical, Users
Featured article statistics Some basic statistics on Featured articles. Monthly January 2004 Yes Yes Yes Articles
Good article statistics Some basic statistics on Good articles. February 2007 November 2005 Yes Yes Yes Articles
What Google liked Google has a web page called Google Zeitgeist about search patterns and trends for the web in general. This can tell us at Wikipedia what people are looking for. Do we have content for them to find? If not, it would be good for us to have as high priority the creating of such a listing, especially for the most recent top ten searches. Ideally, the Google Zeitgeist stats should match the Google to Wikipedia links stats. See also Wikipedia:Articles which are number one for one word Google searches and Wikipedia:Zeitgeist (2004 data) March 7, 2003 November 2001 Yes Yes No Users, Technical
List of Wikipedians by number of edits Name is self-explanatory. May 2008 16 June 2004 A little Yes No Editors
List of Wikipedians by number of recent edits Name is self-explanatory. May 2008 May 2004 No Yes No Editors
Milestone statistics Languages (dates milestones (defined as number of articles in given Wikipedia) reached, in order of reaching them) February 2007 None, but history shows updates from Nov 2004 No Yes No Articles
Modelling Wikipedia's growth This page analyses the article count data in Wikipedia:size of Wikipedia and attempts to fit a simple numerical model of past and future growth to the observed article count size and growth data. November 2006 June 2003 if you want to dig through history Yes No Yes Articles
Most referenced articles These are the most referenced articles as found in the database dump of January 25, 2006. January 25, 2006 August 14, 2003 in page history Yes Yes No Articles
Most frequently edited pages Obvious. May 2008 January 2004 No Yes No Articles
Most popular pages October 2001 Obvious and not updated. See Popular pages. October 2001 No Yes Yes No Articles
Most-edited talk pages Here are the talk pages with the most revisions, as of November 11, 2003. November 11, 2003 February 2003 No Yes No Articles
Growth of Wikipedians by language. Many pages in the category, particularly interesting: Wikipedia:Multilingual monthly statistics (panorama) and Wikipedia:Multilingual statistics. February 2007 July 2001 Sometimes Yes No Articles
People by year Uses birth and death categories to count number of articles about people born/dead in a given year. See also Wikipedia:People by year/Reports. July 2005 September 2004 No Yes No Articles, Technical
Pools Pools have been created in which people make guesses about various future milestones for Wikipedia, with milestones defined as 'when will Wikipedia reach x-number of articles'. May be useful for some prediction analysis. Various Various N/A Yes No
Popular pages A list of pages ordered by number of views in recent month May 2008 April 2004 No Yes No Articles
Wikipedia:Productivity of Wikipedia Authors Activity of editors per language of Wikipedia mid-2006 No Yes Yes No Editors
Researching Wikipedia This page discusses some ways to quantitatively measure our success with Wikipedia. Basically an essay about Wikipedia statistics. 2003 No Yes No No All
Search engine statistics It records data about the frequency and prominence with which Wikipedia appears in search engines (Google). November 2005 No Yes Yes Yes Articles, Technical
Size comparisons This article compares the size of Wikipedia with other encyclopedias and information collections. February 2007 September 2002 in article's history Yes Yes No Articles, Users
Size of Wikipedia Old statistics page. Mostly historical. Some up to date, some not. Check history. Yes Yes Yes Articles, Users
Wikipedia:Statistics The main official statistics page. Mostly up to date. December 2001, but nothing useful there. Yes No No All
Stub percentages With Wikipedia crossing one million articles in early 2006, I asked a simple question: what proportion of those articles are stubs? July 2005 No Yes No Yes Articles
Wikimania 2006 Wikipedian Survey A small survey on the reasons behind the success of Wikipedia. Open ended questions: What drives people to edit Wikipedia in the first place? Why do editors stay with the project? What has editing Wikipedia given you in return? Anything else that you would like to add? How old are you? How often do you edit? What is your highest user level (anonymous, registered user, admin, bureaucrat, steward, developer, board member, jimbo)? Summer 2006 Not repeated Yes No No Editors
Wikipedia interwiki and specialized knowledge test How much more information there is for Wikipedia to assimilate? 22 July 2006 Not updated Yes No No Articles
Requested Articles Bot stats This page shows the current number of requests on each of the requested article pages that the RABot can process. Also shown is the max/min number of requests that have been observed on each page since the bot started running and the number of completed requests that the RABot has removed. The "per day" figures reflect the number of days that RABot has been used as an aid on each page, which may be less than total number of days the script has existed. Initial cleaning, including the hundreds of requests removed the first time this was run, are not included in these totals. June 2006 June 2005 Yes Yes No Articles
Wikipedia:Statistics Department This project, the Statistics Department, provides a space for contributors interested in statistics to discuss what to measure when, and how. inactive inactive some no no
Words per article One of the metrics in the Wikipedia:Size comparisons page is the number of words per article. Some Wikipedians anticipate the rate of new article creation eventually slowing down, and effort going instead to improve the quality of existing articles. This page examines a couple of trends loosely associated with quality: the number of words per article, and the number of revisions per article. October 2005 January 2001 Yes No Yes Articles
Does Wikipedia traffic obey Zipf's law? Zipf's law September 2006 No Yes No Yes Users
Wikipedia:Xiong's stats This is a preliminary analysis of selected English Wikipedia statistics over the period from 2002 January to 2005 March. Data is examined for evidence of a shift in Wikipedian Community values and cultural makeup. 2005 March 2002 January Yes No Yes Articles, Users, Editors
Wikipedia:Traffic Some late 2002 / early 2003 daily traffic figures for the English-language Wikipedia in hits/day: 2003 2002 Yes Yes Yes Users
WikiProject creation trends Using WikiProjects-related meta-data as a window on Wikipedia evolution. July 2005 None Yes No Yes Editors, Articles

Category description:

This category aims to include resources for researchers in two capacities:

  1. . Using Wikipedia as a research tool (see Wikipedia:Researching with Wikipedia)
  2. . About Wikipedia as a research subject (see meta:Research)

We are interested in the 2nd subcategory which surprisingly has very few pages.

Item Description and comments
Wikipedia:WikiProject Wikidemia This project, Wikidemia, provides a space for articles related to academic research about Wikipedia. Semi-active. Wikipedia:Wikipediology is a forgotten inactive version, it appears.
Wikipedia:Academic studies of Wikipedia An incomplete list of academic presentations and papers on Wikipedia.
Wikipedia:User survey Forgotten proposal, see meta:General User Survey for a little more advanced, unfortunately also inactive. See also Wikipedia:University of Würzburg survey, 2005.
Wikipedia:Researching with Wikipedia While it is a resource for the first category, it is a good article and a good introduction to Wikipedia from a more academic perspective.

The following tools are useful for research/stats analysis of Wikipedia and related projects.

Item Description and comments Keywords
API query This API provides a way for your applications to query data directly from the MediaWiki servers. One or more pieces of information about the site and/or a given list of pages can be retrieved. Information may be returned in either a machine (xml, json, php, yaml, wddx) or a human readable format. More than one piece of information may be requested with a single query.
IBM History Flow tool A nice tool from 2004 (download) that led to this article, unfortunately there is no (known to me) 'how-to' use it, and it was designed for pre-1.5 MediaWiki (SQL-based), meaning it may be mostly worthless now. If somebody can update and create a sensible 'how to use it', please do.
WhodunitQuery A Windows-based application developed for the English language Wikipedia. With it, the user can load any Wikipedia article, select a certain phrase, and, with one click, it will search through the page's history to determine who added the phrase. May be quite useful for some content analysis.
Edit counters Editcounters. Easiest way to get some useful statistical data on this side of trying to deal with database dumps. Particularly useful: TDS' Article Contribution Counter: list of contributors to the article by number of contributors (lumps anon's together, use this to get a list of anons); Interiot user stats Tool 3 and Tool 1 (different layout – different stats accessible easier in each one). Flcelloguy's Tool – I will test it soon, looks very promising. List of articles created by user.
Scripts I find the following scripts useful for gathering data: History and Edit Summary Use Analysis (useful, but can crash browser from time to time, and description ('codebook...') of some stats it calculates is not very clear), New pages log and New Users log edit counters (haven't tried that yet)
WikiXRay on meta The main goal of this project is to develop a robust and extensible software tool for an in-depth quantitative analysis of the whole Wikipedia project. Looks promising but not very user friendly at the moment (pre-alpha level)).
WikiEvidens WikiEvidens is a statistical and visualization tool for wikis.

See also

edit