This page is a dead proposal from way back in 2005, but there's a live article rating feature courtesy a grant: mw:Article feedback

We've wibbled on the mailing lists and various wiki pages about Wikipedia 1.0 for a year or two now. We now have an actual publisher asking us for a paper version. Below is the closest we have so far to an action plan.

The key to success will be to harness dilettantism and let the wiki do the work. That's what got us this far.

The plan below includes milestone versions 0.7, 0.8, 0.9 and 1.0. Please think of earlier milestones. To get milestone 0.7, we will use the m:Article validation feature as a distributed rating system.

"We learned from Nupedia that excessive a priori formalization is a killer. Better to start from a very open point of view, and then upon review, if our end product is starting to be bad in some way, make adjustments based on what we've learned." [1]

Note that this plan is concerned with the single-volume paper Wikipedia 1.0, not the possible CD/DVD plan. (The main difference is we probably won't need to cut subject areas for a DVD — we have a lot more space.)

(Feel free to edit this. Discussion to the talk page.)

Action plan outline

edit
  • Jimbo has stated that he doesn't want a fork from the current content, except at the last moment before preparing for print.
  • We have various plans for rating articles or versions of articles, which we can use to see what's up to scratch.
  • We have various lists of what should be in a single-volume encyclopedia. Completing these lists would be a fine program for volunteer recruitment, by the way.
  • In a lot of cases, the lead section is going to become the print article. We need stand-alone leads.
  • We have too many articles with images that are fair-use, permission-granted, non-profit or of other non-free status.
  • We have an interested publisher, one who actually gets it, waiting in the wings.
  • We want a paper Wikipedia in every classroom in Africa.

If there are other milestone steps, feel free to add them, or mention them on the talk page.

Milestone Result Work to get there Prerequisites before starting
0.7 A book's worth of sufficiently highly-rated articles. Just set a minimum rating and then pull from the database. A rating system running for a month or two.
0.8 A more balanced selection of coverage. Writing articles of sufficient quality to fill gaps — heavy lifting. A plan of space per topic, so we know where the gaps are. Complete m:List of articles all languages should have.
0.9 Coverage-complete, with the right selection of tolerable-to-good articles. More heavy lifting. More polishing than writing. Fact checking. Enthusiastic determination. Grim determination.
1.0 Polished 0.9. Improve tolerable articles to good. Good writing. Proofreading. Editing. Patching remaining holes.
Paper Wikipedia 1.0 1.0 prepared for print. Prepress work. Precise requirements from the publisher.

Paper Wikipedia will not be a fork

edit

Jimbo has decreed that we are to fork for print at the last second, rather than fork now and polish. I suggest we trust his vision.

However, consensus approaches the idea of rating particularly good versions of specific articles.

Rating systems

edit

Editorial committees — set some poor buggers to rating thousands of articles — produce high quality results, but will not scale. To scale, the rating system will need to be distributed amongst the editors and readers.

For now, we're going with Magnus' m:Article validation feature. We hope to switch this on in MediaWiki some time soon to gather data. (You can try it out on http://test.leuksman.com/ — note that Brion doesn't like the current implementation.) Then we'll release the data for analysis.

Article versions will be rated on various criteria on a scale of 1-4 or whatever. See m:En validation topics.

This is good because it will be easy and people will do it casually. Harnessing dilettantism, that's the wiki way.

Jimbo's idea is to set up rating on a large wiki (en: or de: — note that de: has an active article quality drive) and just gather data for a month or two, doing nothing with it. Then release the data and everyone can evaluate the results for sanity. We can then do something with the data in MediaWiki 1.6 or later.

Neil Harris has written a marvellous outline of machine assistance in article rating [2][dead link] that could work off the human ratings to give estimated ratings for other articles. Probably need some students after a Masters.

Topic selection for a general encyclopedia

edit

m:user:Mathias Schindler has suggested it as a validation topic — see "Encyclopedic general interest". This should show us what areas the community really think are "encyclopedic".

Note: Whatever are the most popular pages on the site will almost certainly have to go in. Including the odd and amusing stuff.

Letting the wiki give us milestone 0.7

edit

The article version rating system will let the wiki do the work for us:

  1. Wait till a lot of articles (or a fair few) have been rated.
  2. Set a cutoff level that gives you a book's worth of articles.
  3. Examine just how imbalanced we are.

This will give us Wikipedia 0.7, let's say. 0.8 can be better, 0.9 can be area-selection-complete, 1.0 can be a polished 0.9.

Step 3 above can be in comparison to what the rating system tells us the Wikipedia community considers encyclopedic topics.

Bringing areas up to scratch will still be real heavy lifting. How much real work, we can't know until we get 0.7.

Stand-alone lead sections

edit

These are currently optional (according to the MoS), but will become very important because for long articles, we may just pull the lead section as the print article. News style and summary style are your friends. I've suggested "Lead section" as a rating topic.

Non-free images

edit

These should be cleared and replaced where at all feasible.

Note, by the way, that the publisher would probably have a much easier time getting copyright clearances for a fixed paper volume than we do for a changeable open-content encyclopedia (e.g. Tube map, where not having the map would be ridiculous) — though I don't expect anyone to agree this is a good idea.

Publisher

edit

We need to know precisely what they require from us, so we know what to aim for.

  • Size in pages, translate to bytes
  • How much space for images
  • Format for publishing

A lot of the prepress gruntwork can likely be done by Wikipedia volunteers for free. Also, we should be able to test automated systems on 0.7 onwards.

A paper Wikipedia in every classroom in Africa

edit

How do we get a list of schools? How many copies is that? How cheaply can we print a usably good book?

We can almost certainly hitch a ride with other charities for the distribution.

See also

edit

1.0-specific (Jimbo)

edit

Selection process

edit

Rating mechanisms

edit

Bringing areas up to scratch

edit