[22:14:29] tgr: Hi [22:14:43] hi smartyrad [22:15:33] tgr: I have followed your suggestions and created a patch for task 4 here https://gerrit.wikimedia.org/r/#/c/383299/ [22:15:54] Kindly do check it out when you find time [22:16:28] will do [22:17:02] tgr: Sure, thanks :) [22:22:05] tgr: I was drafting the proposal for this project. In the timeline section, I divided the project into two main phases. Phase 1 for 5 weeks and phase 2 for 7 weeks [22:22:45] srish_aka_tux: nice quarterly report. :) [22:23:12] Phase 1 will consist of using ml to check whether the user is a bot or not and Phase 2 will be regarding designing a new captcha [22:24:24] smartyrad: I would expect phase 1 to be most of the work [22:25:17] I'll comment on all the proposals today [22:25:50] tgr : Ah I see, I'll finish mine and upload it today [22:26:39] you can just mark it as a draft if you feel it needs more work, that's fine [22:29:42] tgr: For captcha, Questy captcha looks pretty strong to me. I have mentioned in my proposal that we could use a modified version of questy captcha (One that is usable by people with visual impairments as well). Does that sound like a good approach? [22:31:25] tgr: The reason I chose Questy is because we get to define questions. Those questions could be defined in multiple languages as well which will also solve https://phabricator.wikimedia.org/T7309 [22:37:19] smartyrad: experimenting with traditional captchas is not really part of the project, if the bot detection fails, just fall back on what would get used otherwise [22:37:30] on Wikimedia wikis that would be FancyCapthca [22:37:58] QuestyCaptcha is good for small wikis which can write their own questions and spambot writers don't care about them [22:38:21] Wikimedia is an important enough target that spambot writers would try to adapt [22:38:37] which is easy when all your captchas are taken from a small pool [22:40:17] tgr: So you mean to say that we must focus primarily on using ml to tackle bots and as far as captcha is concerned , we just fallback to the default one that is being used instead of creating a new captcha plugin. Am I right? [22:48:09] yes [22:49:24] smartyrad: there have been past attempts with coming up with a more classical captcha, like image recognition, but they haven't gotten very far [22:50:09] in any case that problem would be a separate Outreachy project at the least :) [22:54:47] tgr: That's true. Thanks for your help, I'll modify my proposal accordingly. Infact this line "which could just be a regular captcha, but if we can generate questions based on image recognition or other hard-for-robots-easy-for-humans tasks, even better" in the description of our task is what led me to believe that maybe we have to design our own plugin as well :) [22:56:08] uh yeah, that was written a long time ago [22:56:24] before it was made into an Outreachy project [23:07:06] tgr: Oh I see, that explains it