Last modified: 2012-08-04 21:10:37 UTC
Dario will create a special 'hidden category' (AFT5) of articles that he will select for testing in phase 1.0 and 1.5. The AFT v5 extension will only appear on articles that belong in that special AFT5 category.
Here is Dario's current plan for this article sampling (assuming no geo-cordinates to work with feedback link option B): 1) we generate a list of articles with no geocoordinates and no redirects from the entire enwiki 2) we sample ~106K articles from this list * 3) we add these articles to a newly created Category:AFT5 via a bot 4) we use this category to activate the widget on eligible articles (after processing the AFT blacklist) 5) we end up with a final sample of approximately 100K eligible articles with no geocoordinates that we will use to collect data This plan may change, but I wanted to post it here, as an FYI.
Here is more info on our plan to keep two separate lists of articles, as described below: White List: Article_Feedback_5 Black List: Article_Feedback_Blacklist Once you support these category names in the beta code on prototype, Dario and I plan try to add a few articles on prototype in these categories, so we can test if the system works as intended. _______________________ From Dario: Here's a summary of the AFT activation logic to be used for the 12/12 launch. 1) AFT checks whether an article is in the whitelist: http://en.wikipedia.org/wiki/Category:Article_Feedback_5 (I just made up the name, we can change it to whatever we want as this category will be used for testing purposes only and will disappear once AFT goes in production at 100%) This category will include articles that are "safe" for the test phase, i.e. a random sample of articles taken from enwiki with the exclusion of: • redirect pages • geotagged articles I am generating this list (A) right now and waiting for Mark to confirm whether my method to detect geotagged articles is accurate. I have a second list (B) ready that only excludes redirect pages, should we be able to drop the no-geotag limitation. We will need to run a bot to populate the category with articles from this list, whether it's (A) or (B) depends on OmniTI's progress with the geotag issue Howie can you check whether Kaldari or Asher can help with the bot? 2) AFT checks whether an article is in the blacklist: http://en.wikipedia.org/wiki/Category:Article_Feedback_Blacklist We should honor this blacklist after checking for the whitelist as it includes manually flagged articles on which AFT (whether it's 4 or 5) should be disabled. This list includes in particular disambiguation pages that cannot be detected using page titles and is constantly updated by the community. So to recap, we only display AFT if article X is in Category:Article_Feedback_5 and is not in Category:Article_Feedback_Blacklist Dario
*** Bug 32838 has been marked as a duplicate of this bug. ***
This problem has not been solved to my satisfaction, based on what is on prototype now. For example, this unlisted article should not be displaying AFTv5, because it is not in the 'Article Feedback 5' category: http://prototype.wikimedia.org/release-en/AFT5_unlisted_article Please go through our testing page to verify that all conditions on that page have been met: http://prototype.wikimedia.org/release-en/AFT5_test_hub
Ok, here's the story with displaying unlisted articles: 1. As per current requirement we are to use the existing mechanism for selecting articles 2. However, in the existing implementation, an unlisted article can still "win the lottery" and have AFT displayed. 3. My suggestion: remove the lottery functionality. Please advise a.s.a.p.
Let's discuss during our weekly call, so that I can get Dario in the loop. My understanding is that he will do the lottery on his end, and put it into the Article_Feedback_5 category. So technically, it shouldn't affect OmniTI directly, if I understand it correctly.
In that case we need to remove the lottery code, and re-test.
Let's make sure we're talking about the same 'lottery', to avoid any confusion: 1- random bucketing: we need you to assign a feedback form bucket at random, using a lottery-like process 2- article lottery: Dario will select sample articles using a random lottery process, and put them in the Article_Feedback_5 category, without any intervention from AFTv5 or OmniTI On the second point, if an article is not in that category, it should NEVER display AFTv5 Which one of these lottery processes are we discussing here?
