Wikipedia talk:STiki

Welcome to the talk page for the WP:STiki tool. It works just as any other talk page. However, before you report a bug or feature request, please check the table of the first section to see if the issue is already known. Please do not edit that table manually, instead just start a discussion and STiki's developers will update the table as appropriate.

STICKY: Known Bugs and Feature Requests

STiki Outstanding Bugs and Feature Requests

ID	Contact	Type	Description	Status	Urgency
T#018	Yaris678	Feature Request	Modifications to the warning system. Better automated logic for warning increments, especially as it pertains to month boundaries. Add option for dialogue box before severe warnings distributed and AIV reports sent.	Pending	Medium
T#010	Allens	Feature Request	There is a desire to somehow represent (possibly with underlining or a color) whether a wikilink has an existing article destination or is a "red-link". Some investigation is needed to determine how to best do this with the Mediawiki API, and the performance penalty it causes. Could also be a STiki option whether or not this should be done.	On Hold	Medium-Low
T#028	John_of_Reading	Feature Request	Expand processing to include some alternative namespaces, namely "portal" and "help". Existing classification models would be re-used and there would be an explicit new queue for these edits.	Pending	Medium-Low
T#027	Arc_de_ciel	Feature Request	More careful recording of "pass" actions. This would also permit a feature allowing a user to view, ex post facto, how the edits they passed on were eventually classified.	On Hold	Medium-Low
T#004	West.andrew.g	Feature Request	The STiki client currently talks to the backend using only MySQL. MySQL communication could be firewalled by some organizations, making the move to HTTP/PHP communications a good idea. However, this remains low priority given that such little mention has been made of this shortcoming.	Pending	Low
T#003	Chicocvenancio and Meiskam	Feature Request	STiki exists only in English, and should perhaps put structure in place for localization. Meiskam wrote some initial structure for this, but I am holding on trunk integration. To move forward, we need someone who has the time, expertise, resources (a server), and willingness to run a STiki backend and do all text translation.	On Hold	Very Low
T#999	-	Feature Request	Note that all requests pertaining to the inclusion of back-end classifier features should be lodged and discussed in the sandbox at Wikipedia_talk:STiki/Feature development	-	-

CHANGELOG for the 2013-05-23 release

Greetings STiki users. Below is the CHANGELOG for the 2013_5_23 release (available for download on WP:STiki), wrapping some small fixes that arose during my dissertation writing. As always, thanks for your continued support. West.andrew.g (talk) 05:14, 23 May 2013 (UTC)

• Fixes to the AGF message dialogue (T#025). One can now abandon the revert action after the dialogue has been displayed, and checks ensure that the customized message will not be placed if the revert fails. Both edits (the revert and the message) commit after the message has been submitted.
• The AGF message dialogue has been updated with the modifications made by users at WP:STiki/Good-faith-revert_messages. Similarly, the number of user customizable messages has increased from 2 to 4.
• The space bar is now disabled when the "classification panel" has focus (most of the time). Previously, pressing spacebar would repeat the previous classification action taken. However, this seemed to cause far more accidents than intent-driven uses (T#026).
• The "user talk" links in the interface (in the "metadata" and "last revert" panels) now append the affected page title in a special format that makes welcoming/warning messages easier for Twinkle users. See WT:STiki/Archive_11#Feature_Request:_Article_Reference_on_Talk_Pages
• Minor changes to backend queuing, stopping the CBNG queue from displaying edits made by STiki users.

Other namespaces?

Hi! I read your dissertation after it was mentioned in this week's "Signpost", and that set me thinking.

For over two years now I've been checking the recent changes feed for the "help" and "portal" namespaces. No bots patrol those edits, and I *think* I am the only editor checking them systematically. So when I check in at 6-7am in the morning (I'm in the UK), it's not unusual to find that highly-visible pages have been overwritten hours ago with stuff that ClueBot would have reverted from articles in seconds.

Now, Cluebot won't touch pages outside the main namespace because it hasn't been trained. Could these edits be fed into the STiki queues somehow?

Here are examples of my reverts in the Portal and Help namespaces.

One "gotcha" is that the page view statistics are not reliable for portal sub-pages. For example, the statistics for Portal:Technology/Intro claim that it is viewed very rarely, but it is actually transcluded into Portal:Technology and is viewed 2000 times a day.

Thoughts? -- John of Reading (talk) 08:58, 3 August 2013 (UTC)

Oh goodness, some has actually *read* that? The namespaces you mention seem to be very valid candidates for autonomous inspection. STiki and its metadata model are perhaps better poised to do this than the CBNG folks (I imagine the "help" namespace in particular has some very different language conventions than normal article text). There are some barriers to accuracy and bootstrapping, e.g., the reputation metrics for these articles need to accumulate. The one you mention about page popularity statistics is an insignificant factor, though.

There will be some minor issues and we probably won't be able to reach the performance benchmarks achieved with vandalism. However, if its just a matter of piping these edits through existing models to provide an improvement over the status quo; then this shouldn't be terribly hard from a code perspective. I'd prefer to dump these edits into an alternate queue for inspection. I'll put this on my investigation list (and add to the feature table); but I'll be honest that I am a bit backlogged right now with stuff for WP:Turnitin and some page popularity stuff. Thanks, West.andrew.g (talk) 16:59, 5 August 2013 (UTC)

Yes. This seems like a sensible improvement on the status quo and doing it as a separate queue makes sense.

Would it make sense to also include the Wikipedia: namespace? I think this will be similar to Portal: and Help: in the important respects. Or is the Wikipedia: namespace better covered by other things?

The User: namespace would be an interesting one... I think it would be basically the same but we might want to exclude edits made by a user in their own user space.

Templates might also be similar in terms of the learning from metadata. However, it might be worth putting the Template: namespace in a separate queue, just because some STiki users would be more familiar than others with the template language.

And then you've got all the associated talk pages. I suspect these will be different in some way.

I hope I haven't just made things unnecessarily complicated. You could, of course, just start with the Portal: and Help: namespaces and see how that goes before considering the other possibilities.

Yaris678 (talk) 17:25, 6 August 2013 (UTC)

Anticipated STiki Downtime, 8/16

I have been notified by network administrators that there will planned network downtime on Friday, August 16th, between 7 AM and 10 AM (Eastern time of the United States, i.e., NYC time). This should be a decent opportunity to test the machine's ability to bring itself back online. I will be watching the status and my contact in PHL is also aware of the possibility things might not go as planned (in which case transitioning to Labs infrastructure will gain some priority on my TODO list). Thanks, West.andrew.g (talk) 17:32, 6 August 2013 (UTC)

Evaluation of STiki

I see some editors making hundreds of reverts in a day or two using your tool (mostly edits by IPs, by the way) and I'm left with some questions:

• Have you evaluated your software/script to see how many acts of "vandalism" that you detect are actually vandalism?
• Are you currently evaluating your success rate in detecting edits that deserve to be reverted? Do you spot check examples to see whether edits you highlight as bad or vandalism are not valid?
• Is there some kind of bragging rights that go to editors who complete the most reverts over a short period of time? Do you think quantity of reverts is somehow more important than the amount of thought and consideration that should be given before reverting another editor's work?
• Do you target edits from IP users and are they subjected to more scrutiny than registered users?

69.125.134.86 (talk) 18:16, 10 August 2013 (UTC)