On Tue, Mar 11, 2014 at 3:48 AM, Steven Walling swalling@wikimedia.org wrote:
The hard part is figuring out what edit summaries are so common that they should be canned. Since there are so many different kinds of edits, that's the difficult part.
Indeed! Here's some research data from Oliver Keyes about the canned summaries gadget that's on enwiki:
-------
So, the research question; what does the usage of the default edit summaries gadget look like?
Using data for the last 60 days from enwiki, I investigated.
Results 371 distinct users who edited in that time period have the gadget enabled. This is 0.1% of the editors who have edited in that time period (224,946). 5,392 edits were made by these users that matched the dropdown options, which is 0.05% of the total (9,145,360). One counter to this is that most people are unaware of the tool's existence, which is true, but edits were made in the last 60 days by 2,338 people who have it enabled. In other words, only 15% of the people who even have it enabled find an excuse to use it in 2 months.
In terms of how this usage was distributed, I've attached two graphs. One is the distribution of those edit summaries over users - in other words, how many users were distinctly using each one. The other is the distribution over edit frequency - how many times they were used, full stop.
Conclusions As a researcher, I cannot find any evidence that this gadget is being widely used, however widely it might or might not be installed. It is sourceable to 0.05% of recent edits, at most (see 'caveats'), from 0.1% of users. Accordingly, I recommend against treating it as a feature for general release without a decent research plan for following up on its usage and investigating how people react to it.
If such a release (and plan) is desired, I'm happy (as Mobile's Secondary) to be involved. It looks like an interesting project, even if there isn't quantitative support for the idea of its utility.
Caveats This is only enwiki; retrieving substantive chunks of the revision table is painful for global queries. If I had more time I could probably have done it trivially. More importantly, the actual edit summary strings are presumably localised on a per-language basis, and so the query would have to be tweaked on a per-language basis - and those strings live in a .JS file, which isn't easily accessible in an automated fashion.