 So I work in the discovery department specifically on search. We do a couple of different things, but I'm in the search area of things. One of the big difficulties in improving search results is you need a way to measure your changes to the search results. And there is a lot of research into this area in information retrieval. A couple of the ones that you commonly are on screen here, they're mean average precision and discounted cumulative gain. A special thing about these is that you need labeled search results. You need results that are scored as relevant or not relevant or some range in between to run these algorithms to compare if your search results got better or not. So one of the things that I've been working on, just too much spare time, it's essentially a search result labeler. Right now this isn't in production range, just runs locally and uses... But anyways, you can log in by a media wiki account. It'll go over to the thing, it'll ask you if you want to log in, say yes and it sends you back and then it just drops you into a thing where you get results that have been previously imported that you can then rate. So we could say for this one, Elizabeth the second, is that relevant for clean? We could go with probably relevant, I don't know, but you just keep clicking on these links and it'll save the score that you give and it will just keep giving you back results that you haven't scored before until you run out of things to score. You can suggest queries to be imported. So we could say we want to add mean average precision and we will import 10 results from English, in this case English Wikipedia because that's the one selected. Those will be imported and then anybody who is scoring results will start seeing those showing up in their things. It's relatively easy, something I think we can get users involved in helping us to build up scores for what is relevant and what is not relevant. There are caveats here because we only import the top 10 results. If something is relevant, but it's not actually in the first page of results, then we don't have it here. Things to think about, but overall I think it's going to be a useful addition to we're calling it Relevance Lab. It's a thing where we can run sets of queries and measure the changes to those queries and try and improve the results that we return for those queries. I guess one of the things we can look at, it can then dump out the scores for you. In this case I'm the only user so everything just has one score, but you can see for the query what's relevant, what's not relevant. You can see from Mount Rushmore, a bunch of radio stations that are not relevant. Then we can start using those scores in our Relevance Lab to optimize for what we return in search. That's about it. Any questions? For now if people could take any questions that they have and add them to the ether pad for Eric so he can take a look at them, that probably would be the best way to do it. We can do it somewhat asynchronously. And as a reminder for anybody who is in the room or is on BlueJeans or wants to be in the room or on BlueJeans you can sign up to do your demos on the ether pad called Credit. That's uppercase C-R-E-D-I-T. Very cool, Eric. Very cool stuff. I think up next is Stas and Eric if you could turn off your video once you're done. Stas, you ready? Yeah. Let's see. Can you see the screen? Yes. Okay, great. I wanted to talk about the Wicked Data Query Service GUI and the new stuff that has been done since basically for the last couple of months. So basically Wicked Data Query Service is the query interface for Wicked Data which allows to ask all kinds of queries. And we have this GUI that allows you to do some interesting things with it. So first of all, this is the interface. So that is for like people just beginning with there is a lot of examples which allow you to see how to do queries. Like the first one is the list of cats on Wikipedia which has, so Wikipedia has, Wicked Data has 59 cats. There should be more. So one interesting thing. So if you see the query here it's kind of not very comprehensible what that means. So if you actually go there and see the tooltip it explains basically what these things are. So this one means instance of and this one means cat. So you can understand this query more. So now this is for the results. So what you can do with this results you can look at them. You can click through to Wicked Data and see the actual entry for this. You can download it in various formats and you can have a link. So if you click this you will get a short link to this query. So if you send somebody this link you can go to this page and have this query displayed for them. So the next thing is also when you edit the query. So if you didn't notice there is also auto-completion. So if you type, if you want to type, you just type instance and it substitutes. So now I want a cat. So you can see if you want instead a dog and you can replace it with a dog and the same works with a label, with variables. So you can figure out all this without remembering all these numbers. So there is also some improvements in the UI. By the way, you can search this. So there is a query for paintings. Paintings by Gustav Klimt. So you can see that this is painting and the creator is Klimt. So if you run this query you got a bunch of images. So this is not very useful just the file name. But what you can do if you click on the picture you can actually see it and browse the actual paintings. Even better, what you can do is the image grid and all those images are displayed to you as a grid so you can see them all. Okay, what else interesting? So you can also if you want to see these examples in Wikipedia. So there is a page in Wikipedia that has all these examples and the code behind them. So if you have some example that you feel particularly good about you are welcome to go and put it on this page and this would be automatically displayed in the query examples and will be searchable and available for everybody to see. There are also, so there are menus that allow you to see some interesting things like for example there is a properties browser. So if you want to know what properties there are and for items you can look through it and you can search like what properties are related to countries so you can do a search and see which properties are related to countries. So what you can do is so what else and there is of course link to help that you can read if you need some details about this and you can provide feedback to this. So one thing that is being worked on also is coordinates so if we take the example of let me see if I can find it. So for example this one is list of bridges so you see there is a coordinate but you cannot really use it very much because it's just numbers so what is being worked on is to hook it up to actual map and kind of displaying this on the map. So this is basically all I wanted to show so we are working on kind of improving them if you have any suggestions of what else we could do and how we could improve this tool to kind of make it easier to work with it you're welcome to put it on the discussion pages or on the feedback page. So that's it so if you have any questions please tell me. Okay also cool thanks a lot Yuri I think you're up next. Hi everyone I'm having a lot of issues with blue jeans it has crashed my computer just now I'm very happy about it I'll hopefully be able to present though let me share my screen. So I'm going to take off right where it does and I'm going to show you what it looks like which is sparkle sparkle is wonderful sparkle gives us all this wonderful data well this is a way to visualize the data remember there's your coordinates well right here we have a graph extension graph based on a graph extension that plots the disaster is the biggest so we'll be able to do this kind of things and we actually you already can do that for example if the graph is static you can already include it in Wikipedia to pull the data directly from Wikidata that's one thing thing number two someone from Vega team which is the basis of the graphs has contributed this wonderful graph about most expensive painting here you seem to have cut out and you get to see some information about the painting as well as you can just filter the graph and say oh just show me a certain period or certain author and that's what it does instead of looking at this giant table which is obviously more informative but not as fun to look at exciting news next week we plan to release maps cartographer extension maps frame map data map link to you'll be able to edit map existing map and even draw on it say like oh I want to square and you draw a square this is coming to the key voyages next week lastly with the graph extension I did some some experimentations with page views API we love it and created these two graphs about like US elections in real time who is view how many people are viewing different candidates pages and also how what is the total viewership of different wiki sites this Wikipedia there's a wiki voyages there's other wikis for the past half a year and I'm done thanks Harry looks like Ron is up next yeah I'm going to route Ron's video and I'm muting my microphone helps too I sound a bit echoey is that me yeah yeah I'll turn you down a little bit I'll bring up a little bit okay how about now sounds good sounds better alright so what I'm going to show you is very rough and ready but we in notifications we have something called bundled notifications which is a feature whereby when for example when you get multiple notifications in the same type for example multiple people edit your user's hot page they end up getting bundled into one so instead of getting a notification saying Alice edit your user's hot page and Bob edit your user's hot page and Alice edit it again you don't get four of those you just get a notification that says Alice and one other person edit your user's hot page but that way you don't necessarily know exactly what happened so we want to make it a little bit better you have some designs for this but what I'm going to show you is a very quick and dirty and rough and ready implementation that I made where this is an incognito window I'm an anonymous user editing catbooks of your hot page and I'm going to write something and when I do this catroll will get a notification and it will show all the details of what I just wrote it will say the section name it will give you the content it will tell you who did it etc but if I then a different user comes in and responds to this comment then now I will not have as useful of a notification it will just say two new messages on your hot page in fact I could even edit a different section three messages on my user's hot page and I won't really have any useful information about what happened now you can probably see which part of it is not ready for production part it's the part that is incorrectly has zero messages and it's halfway through the line but with this code you can expand this and what just happened if you can manage to stay logged in despite MediaWiki's session management software then you can expand this and you can see the individual notifications that constitute this bundle there are many things that don't work yet for example this icon here is misaligned this expand thing doesn't work well with the line here and the G kind of gets bisected the number isn't right initially and you should be able to individually mark these notifications as red which right now you're not because it just masters on the oldest notification to do everything so there's lots of things that I still need to clean up about this but this is one of the things that we're looking at doing next okay awesome well I think we don't have any more demos so we can cut it short unless somebody wants to jump in we've got nine seconds okay that's it there actually is something Caldari's coming he thought it was later but he's coming in right now and he did want to do something okay yeah sorry I was at lunch because I'm two time zones off but let's see if I can get a little demo going let's see if this works real quick bear with me for just a second it's going to load anyway what I was going to show is the new gadget manager that mostly Lego I had worked on a few years ago and Timo worked on some and I think Ron even did some work on it but the community tech team has kind of picked it up recently and been trying to push it out the door and actually get it released so we've been doing a lot of work on that I was going to show it to you on our labs instance I don't know if you're ever to load if I can't get to load I might just have to give you guys the URL and have you guys take a look at it just pasted that's the gadget manager it basically just gives you a really nice interface for dealing with gadgets where you don't have to mess with creating these obscure definitions and wiki tags you actually have a nice graphical user interface it's currently made with jQuery for UI because most of it was written a long time ago but regardless it's still a nice client side user interface for building and managing gadgets and all their dependencies another really nice thing is that it also lets you use midi wiki's i18n system to provide localization gadgets which is a real pain in the ass to do right now the few gadgets that actually do any kind of localization all do it completely differently well anyway for some reason that URL I just got a 503 error when I tried to load it for some reason it says service unavailable so I cannot do a demo I'm very sorry but maybe you guys can try it later on your own with that URL anyway sorry anti-climactic see ya no problem I pasted the link into the etherpad for people who are going to look later does anybody else have anything that they want to demo?