 the showcase of the hackathon during the pre-conference. There was a two-day hackathon where developers on software within the Wikimedia universe and people that aspire to work on it come together and have conversations and do actual work. I'll just, is this better? And, yeah, it's, I think, for the third or fourth year now that we do two-minute presentations during a one-hour time slot in the main conference so that in understandable language hopefully developers will inform also non-technical people of what is being accomplished here and why the heck are we together eating pizza in a room for two days. We have 17 presentations, so at the moment we expect to have about 45 minutes of content for you, which is well within our available time, and I would like to invite, because we could not find Saddam, Adam and or Lego on stage. If your session is about to be up make sure that you're in the front of the room so that the switches go quickly. The project Adam and I worked on is for the winter extension, so if you're not familiar with it, the winter extension was designed to identify errors in Wikitext, so we want to progressively like remove a bunch of the edge cases and historical accidents that exist in Wikitext to make it a standardized language that works for both humans and the parser, so throughout that we've been cleaning up some edge cases, and so the winter extension what it does is it identifies these kinds of edge cases that we want to break in the future, and then it goes through every wiki and it creates lists of articles that editors can go through and fix. So one of the things and this was deployed about a year or two years ago and one of the things I'd wanted to do at the time was like show editors how much progress they'd been making, so we had been recording stats for the past two years on how many errors were left you know like every day, but it was just going into this black box and it was never exposed, but so now what Adam did and me was we created a graph so now once the patch gets merged and deployed there'll be a chart showing editors progress of how many errors they cleaned up, so this is the CEB Wikipedia that's where the data is from and it was there's an error where table tags should are like nested in the wrong way so they get deleted and Adam is going to explain how we did it. Oh so I actually don't know about that but yeah but how does it get there? He does know how it works he's just tired every time like every like 500 edits we save the number of we save the number of errors that are existent in the system called Statsody which is just it measures counts over time and so then it pulls it over the API and then turns it into a beautiful graph. Thank you. Ursula are you in the room? Uh-oh is it option command? How do you make a new incognito? New tab? Yeah no incognito. This is already incognito. Oh it is? Okay that works. Thank you. All right everybody I'm going to show you a few my name is Aaron Halfaker I'm a researcher at the Wikimedia Foundation I run this machine learning team I'm going to be showing you some gadgets that I made that show off some of the things that we've been doing with machine learning and as I do when I do these things I'm logging in with my uh my account and so I'm going to ask are there any uh Wikipedia administrators in the room? I can't see you raising your hand I'm just going to assume that people are uh are there any any administrators in the room who have not enabled two-factor authentication? Uh shame I saw at least one hand and so you really should it helps you deal with situations like this where Sebron might steal my password and he won't be able to get into my account because I have secret numbers that get generated. Come on oh my gosh it worked the first time okay let me get these open is it is control option in max how does this work again control click no option click uh option yeah option click whoa what is this sorry I only use open source software this is super weird uh option click there we go okay so three gadgets one of them I'm going to show you I made this one actually for Rosie step Rosie step uh told me that she was using our article quality model to see how she was improving the art the quality of articles in Wikipedia and it was a pain in the ass because you had to go back to our machine learning tool in order to uh do stuff with this and maybe how do you make it bigger 30 seconds oh come on how does this work option plus option option plus is not working option shift plus that's that's command but anyway what this does is it takes our article quality predictions and it sticks it on you got 15 seconds extra and so you can like scroll maybe if max can do that um nope they can't so we're moving on to the next one which is in basque wikipedia they asked me oh we're really going to need to scroll for this one maybe I can just go to the section there we go in basque wikipedia they asked me if I could have our our article quality predictions there show up on the list of articles and so these article quality assessments that are actually to the left of these links are actually coming from our orders prediction model and of course it wouldn't be complete if I didn't also have the orders prediction model throwing article quality predictions on the top of articles there so nobody really actually has to go through all these articles in basque wikipedia labeling things orders can make some predictions and that'll help people find the important things to work on that are low quality thank you erin next time try to say less next would be rogue assassin one two three or media seven increase the volume of the audio of the computer I think it's uh I think it's been made to work okay so uh being born in the youtube age I like to consume my information through videos and as you all know wikipedia is predominantly a text based encyclopedia so the goal of building video wiki was to visualize the sum of all human knowledge just to show you why video wiki is awesome and what insights that can give it give to you is okay okay so if you see the equivalent wikipedia article or if you go to his early childhood where's early childhood yeah this is one very interesting line here which says at the yeah he thought himself computer programming at the age uh himself computer programming at the age of 12 sold the code of a basic based video game he created called blaster to a magazine called pc an office tech for approximately 500 dollars so now it's a very simple line easy to comprehend uh Elon musk uh created a game called blaster and he sold it for 500 dollars but the same line if you read it in wikipedia a video wiki uh it looks something like this early childhood I'm just getting there so this is the actual game which uh Elon musk has created so in 1980 so you actually get to see the game which he created and this is the kind of perspective what you see in video wiki so that's what video wiki is all about you can it's live on video wiki.org you can play around with it and one big thing about video any video editing website is that it's uh the editor has something known as a timeline feature a youtube editor any editor which makes editing really hard and complex but when your mission is sorry when your mission is that anyone can edit the sum of all human knowledge your learning curve should be really small so if you want to edit this your time is up oh yeah you can just drag and drop it thank you uh sudden you're next um yeah I think you have to go back back back yeah it is yeah you can't see anyone in the room either first one yeah hi everyone so um one of the things given that a lot of our traffic is going to mobile and especially for wiki voyage it's a mobile site this is the current state of the mobile main page on wiki voyage um so I've been working to do a redesign so that we can actually ab test the designs of a mobile main page and so this is the new mobile main page um it's got a box that basically takes you to the search function that's been stolen from some of the work that engineers have been doing on the hindi wikipedia um and come on go faster thank you um and the carousel's been redone so we've taken uh the bootstrap carousel it's uh done by github um so that works much better on mobile and yeah everything's tried to kind of be kept roughly in style with um ui and stuff like that yeah that's pretty much it thank you thank you sadden uh then we have media seven on the pay gap in you cannot use your own laptop you next time close all the windows this will be kept for eternity in the in the live stream and I'm presenting a project as to work on it and it's about the pay gap in Germany so how men and women earn differently the idea was to make a play a game so it starts here and you can relate what somebody earns regarding to the profession so you can move cards like this when it's right when it's right it stops yeah and then you can do it again or you can move to the next step so I'm not going through all of that I'll just show the last one because also you can look in and then decide what you think would be good so it has a lock in button and then you can select how would you make it if you would be um the director of a city how would you how would you um give the money to what profession and to what people so with that it's possible to collect data so and I thought but this is just at the beginning I thought it would be good to do this with wiki data to really collect data from professions and then to see how can we improve that because in that sense it's also political to change the situation of different professions because the topic concerns not only women it also concerns men because in a lot of professions where women work the payment is very bad so also the men who work in that profession are paid very bad and by making this visible um it's possible to change the situation okay so um I put the website on the fabric on the side today um so if you have more question please come and ask me that's for now thank you thank you next is lina on clustering comments categories you still know the password so um for anyone who ever edited commands you know that it's complicated to maintain good categories because you have um several uh of string of images it can be from glam partnerships or people visiting museum so you end up with things like that where you have several images for the same for the same artwork so my idea was how can we regroup them under one category that is going to be linked to the item of wiki data so I first use um I going to show you so first idea is that the information on commands is always going to be um fuzzy so we will not be able to rely on just machine analysis we are going to always need human control so um what do you copy past and this command I think was it command yeah command c and command v so it's going to um so I'm going to load the data from commands since I already did it in cache is going to be more quick than what is used with you also we have this interface where there is a part of clustering that is already done so I can say that this is not a right image so I can move it on the side and then yeah let's say that there are three images that are from the same artwork so I move everything here and now I have two clusters that are represent to your two artworks and I can put it all on commands and wiki data it's going to be uh take us arrived so I'm just going to show you the end result is something like that where you are at the gallery for a category which with each subcategory being an artwork and if you go to see the category of the artwork there's all the categories of the images that are storing the same your time is up and there is also the wiki data item with information we're looking for Eugene who worked together with James here are they here any of them uh dutch keyboard yeah kind of is the is it in the is your uh okay hello I'm Ed Sanders I'm the lead developer of visual editor and as is tradition we like to present our collab pad at every hackathon because we slowly work on it in our spare time and so between each hackathon not a lot changes but we are getting close to the point where we reckon we'll be able to deploy something for people to um play around with in public so this is what it looks like and then someone in a secret location can join the pad oh look there's Rowan oh look he's typing he could be miles away he's actually in the corner there crouched out um as it's a hackathon I thought I'd add some more features so I added this checklist feature so you can do checklists whilst you're on your oh look Rowan just uncheck something I know we've got to check that now because that's done um yeah so we're getting close oh we've got two people on the document now and I can type as well um we have a project page please look on that we had a session yeah yesterday or Friday I can't remember uh where we had lots of great ideas about use cases for this one of which was to take this completely offline so you know kids in um townships can like play around and learn how to edit Wikipedia in a much more fun environment I mean we're having fun and we're supposed to be grown adults so um but also just uh um also we're having a meet-up at 6 p.m in and I forgot the name of the room 30 seconds it's Sino Lario so come join us for a meet-up and talk about what features you like to see or where you'd like to see it deployed or um yes all sorts of great things how much time have I got 15 I don't know how many people have joined dragons oh my god all right someone else can have my time because I'm now just fishing up thanks Ab a few people registered but we cannot locate them Tony Thomas we're looking for you um Andra all right then you're up I should have bookmarked this that looks about right okay then there's still something wrong this of course no no clock is not running so don't worry dash who thought of this naming scheme did we at least we got it beginning in two minutes okay thank you um bit of history three years ago we loaded the complete disease ontology into wiki data the reason is we're trying to combine biomedical data in different contexts and the disease ontology is one of the backbones of loading different medical and biological resources and to answer to do research when we uh loaded the disease ontology in wiki data there was a surprising effect we had suddenly it's an English only ontology suddenly we had on average eight translations for all um all wikipedia of all diseases in disease ontology so at a hackathon I thought about how is it with translations in all the I learned they're not all but 10 of the 11th South African languages and here you see the distribution there is actually only one disease being translated into all uh into all the languages covered on wikipedia so what I did is first at a hackathon wrote this query and give me a second because it might take some time to run and then share the tweet that you just said and ask the whole community please add your if there are empty blocks in your in the to try to add some uh results uh reached out to the chapter the local chapter and they gained some terms so young the hackathon we have increased the numbers uh quite a bit there is 20 seconds one 10 so I was hoping at the end of wikipedia of wikipedia let's make more than 10 but as you can see there are more fives more fours and more sixes that was it so please let's get to the 10 before the wikipedia ends thank you thank you andra uh avan avan ross is next i i'm i'm i'm i'm i'm i'm i'm i'm i'm i'm i'm i edited the template wizard it's a new extension and edited there are some wiki data features so i have two goals one to let editors in wikipedia access to wiki data more easily when they edit edit box info boxes and the secret goal is to have a declarative mapping between a template parameters and wiki data properties so i will show you how template wizard looks like so it adds a new button in the in the editing so you can add a new info box or other templates it opens a dialogue where you can fill up the parameters of the template and this is the template wizard with no wiki data now we want to add wiki data so we add to the template data a declarative mapping between the property in wiki data to parameters in the template we do it only once for the template and then when someone adds the template to a new page it indicated the parameters comes from wiki data and you can see the the data that comes from wiki data you can also click on the wiki data icon to edit the parameter directly in wiki data that's it you can see the fabricator task if you have any suggestions thank you thank you c scott yep next is c scott um so in may 2017 i attended an annotations conference and uh one of the conference attendees a ux designer from npr i contributed this design for structured image annotations um this first image shows uh point we're going to go left to right shows how uh commons looks now um and the second image is this demonstrates mobile aware design of that every annotation has a minimum target size which isn't strictly related to this hackathon um this shows when you actually go to add an annotation uh you initially get a plain text box uh much like you do now but the cool part is when you start to type you get a wiki data drop down and so that's not just a cathedral it's q 43 282 with a bunch of metadata including translations into every language um and then when you select it it describes the relationship you also see that the cathedral says place um uh so it describes some additional hints as to what it is so this is what i actually did in previous hackathons i actually implemented all the back end stuff that makes uh this this generate nice w3c standard annotations but none of it was visible so now i finally did a little bit of visible stuff it still kind of looks scrappy but you can actually see it this time um so um you had the final file annotation is this playing yes um and yes this ui sucks but it actually sort of works um and uh yes very slowly pick it um again when you type it uh displays an entity so that's q 42 and then you pick a type so this isn't just an image it's an image of the grave you save it and wow that's a wiki data statement which is structured a machine readable all sorts of wonderful things uh one minor little fun thing that i discovered the cathedral i'd like to say it's a place or a building but my choices were cathedral or church or ether north church which is a little too specific i added building we'll see how long that lasts douglas adams i can only describe him as human um which seems a little blot like um i'd like to say he's a person but uh it's locked so i can't so we'll see okay thanks great timing thank you next is uh kunal on the way to plus two so one of the things i'd been talking with different people was um um about um onboarding and mentoring new people and i was talking with roan and roan pointed out that we have great um like mentorship and outreach programs for new contributors but for contributors who are in the middle we don't really have anything that great um and these are the people who you know maybe aren't maintainers yet but they're kind of still in the middle like they've been submitting good patches they've started doing a little bit of code review um so what i did was i wrote a draft of a document called road to plus two and it's basically like i want it to be a guide of um of like the things you should start doing you know like how to start reviewing code well um and like areas that you can get involved in um so that way people you know they have um an avenue of like how they can move up and as well like at the bottom um there's like a work with a mentor section so these are the people i've volunteered for this already um and if you're interested i'd love to see like people start editing this document um as well as like signing up to mentor so that way we can um you know like start mentoring the people who are in the middle of the road um yeah thank you thank you uh we are looking for angel obrejon and ishara ah okay good hi this one okay this is a boat after the hackathon the barcelona hackathon i finished my my boat this one only changed the i don't know here for example this one this one is a person is a human is male this is country from spain and also occupation is journalist okay then uh the boat changed the description in spanish to say the spanish journalist and the description in spanish please uh here hey where is the spanish okay this is periodista espanol this is in spanish but now only changed in spanish then uh here i i spoke with my friends in asturian in galician in basque also but i need more language then if you can help me all right thank you very much thank you uh ishara yeah okay hi i'm ishara i was making terrible logos for people's projects on request and um well here's one that somebody apparently is actually using so the wikibase registry is using that logo there see they're using it i don't know why um there were a bunch of others so surprisingly few were actually requested this time which meant i was able to catch up a bit on the backpack vlog from last year but um par suede apparently i made one last year and forgot so i made another one this year and this is the new one uh may or may not be used i would hope to get a shirt with both of them on it just because the previous one was hilarious uh here's another one it's it probably needs a new name but now they have a graphic used for it which is great yeah but the best thing oh here's another one that's um probably okay oh no that's awful sorry uh no no no the best logo here is um this one it's a new logo for media wiki.org and i intend to foist it upon everybody somehow i'm not sure how um in the meantime we also have a very nice version of it that we can use for joke contexts just because this seems slightly more accurate there is also a version of this specific to skinning because quite frankly we need something to represent our um front end interfaces and let's face it media wiki is horrible so it not actually being able to put its pants on seems slightly accurate considering we have no idea how to put skins on it so yeah this is our future thank you for that adam you're up so following on from the terrible logo that we're using um the wiki base registry we haven't really done anything on it this hackathon other than create some new um add some more wiki bases to it uh so if you know wiki apiary or however you meant to say it this is the wiki base equivalent of it currently and maybe my long-term plan over the next year is to merge the two um yeah and so we have this cool new logo um in the run up to wikimania we created two new wiki bases and on the front page there is a link that should work and we'll take you to the query service for the wiki base registry you can run on this very simple query and it will present to you this lovely timeline and if i scroll down uh huh uh in the bottom right hand corner then you can see the two lovely wiki bases that we created uh for people at wikimania and then also the wiki base registry logo by asara and that's it thanks adam next up is uh chase charles mathews are you here good that's incredibly bright uh i don't have any visual cues so imaginations only closing your eyes is optional uh we deployed um an anti-vandalism extension for fabricator oh all right let me get closer uh spam and fabricator has been a big problem most people in this room probably know um mcundin modell has worked on an anti-vandalism extension which is all private for now um and i helped him deploy that and we iterated on a bit the heuristics for automated user blocking need work uh because it forcibly logged out the wrong users pretty quickly um but the one thing we wanted to socialize is there's a trusted contributors project uh in fabricator that's essentially whitelisted and the story with that is if you're a member of trusted contributors you can add others and there's only like 200 some odd people in that group now so we're hoping that that can kind of grow organically um so that the potential for uh adverse reactions is smaller um and then along with that we've been doing a lot more thinking about kind of cross-platform anti-vandalism uh efforts right now we have a bunch of administrative tooling and we've had some bad actors who kind of go from one to the other from garret to fabricator to wiki tech and a lot of times they're starting on the main wikis um so i'd really like to talk to maybe some stewards more people about kind of collaboration and coordination maybe tooling maybe only social that's all thanks uh next is uh yeah mr matthews dr matthews hello let me know if you need help the sound source project so that's wd dot sound source focus list now while i was here and this is not tech i decided to repurpose this from just not just the starting point of our sound source downloading project but actually to be a compact bibliography open access biomedical literature that addresses the systematic biases of the well known of the current medical literature systematic bias is very simple diseases rich people in rich countries get more attention um but in order to compile this list we're using a p5008 focus list of a project and um what i did tech wise is on the tech page sorry um which is wt shortcut in order to add the statements that create the um focus list it's uh just p5008 and then the p is q 55 million something which is the identified with the sound source project so the fundamental tech thing is this papers do is digital optical identifies if you could get them somewhere and if you could translate them into q numbers uh for the items on wikidata you'd have a list of q numbers and then you could create statements q number p5008 q 55 million something and you can put them in uh quick statements and you get them tagged so the big deal is translating from where you ever you start to list of do is and then going forward that's so with the help of c scott understood that um searching in the parzoid of say in wikipedia article you could identify do is easily then extract the identifiers but there's a case issue here i've learned from james here that it should be all uppercase for wikidata 30 seconds we get on to this type of sparkle query that um alexa helped me develop with a values statement and then that will produce uh what you want to go on to the um quick statement stage that is it will get you the q numbers you want so uh the rest is routine for quick statements people and we have a batch input um facility down this route if not others for our focus list thank you very much uh uh jan yeah you're next and we have one submission that would be the last one but it's currently incomplete uh someone wiki light okay can you at least add your name in the it's a pad thanks all right this is a very small hack but it's long due in sweden you used to go to the defense forces before you upload or publish an image that is taken from the air and last summer that switched to the national land survey so we had a couple of the templates that were handling this and now we merge them both and the user doesn't need to know uh which authority it's just entered approval id and it shows the right logo and now there's also 14 exceptions where you don't need approval because the land survey thought this was too much to handle all the influx so you can just add what you claim to be the approval and it will show up neatly there's also some error handling like if you try to add an approval id and an exception at the same time so easier for publishing images for from the air from sweden thank you and uh last presentation this one you want to open yeah wiki light is a search engine uh which displays uh wikipedia articles uh for low internet resource people uh uh it's it display only text information and there is no cross links and it's a prototype project and now uh if somebody got any real interest then i can move forward that thing i'll just demo you uh if you type it's cape down the whole thing will came from cape down if you move over the cursor then you will have a small description there and the whole thing is designed in a way that uh it will load as low as possible and um if you put this thing and go for the search it will give you the result from english wikipedia if you click the read more you will get the cape down article without an image it is very useful for low internet people uh so that's it okay thank you okay um that was the showcase um um i have some uh some closing notes um we have uh the etherpad and the etherpad of the contents uh the the contents of the etherpad will be updated in the fabricator task for the hackathon um uh this was um the hackathon of uh wikimania um the next hackathon uh will be in prog in may 2019 uh and of course next year at wikimania in sweden uh in stock home there will be another hackathon and uh um everyone who was interested in uh joining the wikimedia developer community is is welcome uh at at at any of these events uh if you want to get involved in wikimedia technology uh please go to media wiki dot org uh we have made a lot of efforts to make our development uh a developer portal um a lot better than it was um thank you all for being here uh we like a big crowd next year we would like to have um yeah two few seats for uh the people interested uh full house is the is is what we still have to accomplish for this showcase if you like it please uh tell everyone that they should be here um and uh that's it until next year bye bye