 and I am the manager of the analytics team and here the lightning talks are here to give teams in the engineering product organization a chance to demo quarterly goal that they've reached or a significant milestone and we also open lightning talks up to the community to this to show what they've achieved. Now without lightning talks last 10 minutes so well there will be time for questions at the end and if you're local please raise your hand or go to the microphone for questions and if you're remote go to the channel Wikimedia Tech and I will be monitoring the IRC channel and we'll relay your questions. Also Megan will give an announcement two minutes left and one minute left and when your time is up I will jump in. So without further ado our first lightning talk is by Marcel on the Pageview API. Hi Marcel. So yeah I'm Marcel Reforance I'm a developer in the analytics team and I'm gonna talk about the newly released Wikimedia Foundation's Pageview API. Let me share my screen so you can see the presentation. Okay so Pageview API is a long time ago story. For several years people inside the foundation and in the community I've been asking for this API. Lots of people have participated and bringing new ideas having discussions designing the API implementing it and productionizing it and you can see some other names on this list. So how does it look like? What does the API? What is it? I thought the best way to do this would be a quick live demo so please feel free to follow me in your browser by clicking on those links. I'm gonna open them in separate tabs so I can explain it better. The Pageview API is a collection of public URLs that return analytics analytic data about pageviews on Wikimedia sites Wikimedia Weekies. The first link is the root URL. Let me zoom a little bit so you can see better. It's a root URL for the page for the pageview API. It returns the list of endpoints that the API offers. There are three endpoints right now. One is called aggregate second is for article and then top. I'm gonna go for each one of them separately one by one. First one is aggregate. This endpoint returns time series of page view counts for a given project. In this case I've chosen English Wikimedia and also you can specify a couple of filters whether you want to see the whole pageviews or just the pageviews for the desktop page desktop site or the mobile website or mobile app site. Also you can specify whether you can you want to see pageviews generated by human users or machine users. You must also specify the time granularity and the date range. In this case I've chosen the month of October and the results are like this so it's JSON formatted and there's one one item for each day in this case and for each day you get the timestamp and the view count or English Wikipedia desktop desktop site user human user. So here can you see first of October second of October and the pageviews. Now to the second endpoint. The second endpoint is called per article and it's basically the same as the first one with a difference that you can specify the article name. I've chosen Albert Einstein's article in English Wikipedia for the same period and the results are practically the identity but just you get pageviews for your chosen article. Okay now for the third endpoint is it's called top and it returns the top thousand most view articles for a given project. You can also apply some filters and you have to give a time span. I've chosen 31st of October 2015 and if we look at the results we get main page of course is the number one with close to 18 million views. Surge page is number two and it goes on. It's curious that number five in rank is Halloween which makes sense because of the date. And how did it work? So I've made a small diagram. Imagine this is the readers of Wikimedia sites. They access our media ecosystem and hit the caching layer which is varnish. Varnish responds to the requests but at the same time generates a lot of request logs. Those logs are forwarded to Kafka by a program named varnish Kafka. Kafka is our message broker and it helps buffering the request logs until they can be imported into Hadoop. Hadoop is our distributed computing environment for big data and we have a couple jobs in Hadoop that identify among this raw requests, identify the pageviews because not all requests are pageviews. By the way they use the pageview definition created by the research team and after that they aggregate this data into the format expected by the database and insert that into Sandra which is a chosen database and the last step in the chain is REST Base which is the software that powers and serves the pageview API. It queries Cassandra and it provides the endpoint. So what can you do with it? Analytics team has developed this small demo that's called article comparison that can compare two or more articles together and charts the time series in a graph. You can choose the project that you want. You can choose the time range and of course the articles. This is a very naive example with cat and dog but you can add more things. For sure you will find a more interesting example but you can see that this gets suggestions for articles in the UWK API. You can see how it works. You can also choose another wiki of your choice. I'm Catalan so I like Catalan wiki. I can compare typical dishes. You get the point. I think there are a lot more things we can do with it. I'm curious to see how far we can go but that's an example. So play with it and have fun. Next steps. First thing we want to go is fix bugs. People are already using the API and are finding bugs and filing bugs for us so that's what we're going to prioritize but at the same time I want to collect ideas on new features and improvements and we're not going to work on them right now but next quarter we are going to prioritize them and task them and the next thing I want to say is we are having a session in Wikimedia Developer Summit concerning the API. We're going to discuss this a bit more in depth and answer all your questions. Here's a list of links. Everything that I used in this presentation is here plus a couple links on documentation and a couple ways to reach out to the analytics team for any comments or bugs or anything. That's it. Do you have any questions? Yes thank you Marcel. First question is how far back Oracle data? Right now we have data since September 2015 but we are planning to backfill until May this year. This is still ongoing but that's the plan. Thank you. Any other questions? I'm gonna plant a question. Are there any limitations? Example rate limitations or limitations on the API? Yes for sure we haven't tested them yet but so we're planning to find the limits as soon as possible and then react. So does REST pay? Sorry? Does REST pay? Yes REST pay caches and yeah it's a feature of REST pay so and all the URLs are designed to improve caching. Any other questions about that? Thank you. So how do you resolve the task problem? Sorry I couldn't hear the question. Can you repeat it please? The audience is very bad. Is there any difference between the page views on stats.grob and API? Yes page views in stats.grob.se Is that what you meant? Okay page views on stats.grob.se are based on another source of data. We are calculating these page views on the new page view definition on the research team and this definition doesn't count some page views that we counted before as page views so there may be some differences between them. Yes. So honestly it's gonna be very difficult for me to hear the question. That's a good question. I think this discussion and probably the task that we're using for discussions on the session that we're having in Wikimedia Developer Summit or either a separate email thread. Yeah that's a difficult question. Thank you. Yeah just to elaborate a little bit on that so we can update it and in A because the new definition is better at filtering thoughts so the numbers dropped and the same thing would happen to yours numbers if you slightly jump over to this new page view. But we should continue that discussion see if there's anything that can be done to smooth the transition. Well hey Kevin we're having a little bit of an audio technical difficulty if we can just pause the talk real quickly. I'm gonna exit the hangout and re-end. Okay that might alleviate the problem. Right well thank you and let's give a hand to Marcel. So next up is Dimitri with some demos on the Android app. Hey everyone can you hear me? Yes. Okay great. I apologize if my voice is a little stuffy I'm just coming back from a slight cold. But yeah my name is Dimitri I'm a software engineer on the mobile apps team and I'm the product owner for the Android app. Let me just share my screen. Okay so I just wanted to catch everyone up on what's been going on with our Android app because we've been working on some cool new things and continuing to make improvements all the time. The most significant thing to note is that the app is in the process of switching over entirely to using our mobile content service which is built on top of Restbase. So the content service was spearheaded by another Android developer, Bairnd Sitzman, and all the rest of the Android team has been making small contributions here and there to this service. So for anyone who's not familiar with this the goal of the content service is to enable any application to ask for an article title and give back the article broken down by sections not just broken down by sections but also include all kinds of really useful metadata that would normally require making additional network requests or additional expensive processing by the client. So here's an example we asked for the article on Paris and here's the response from the server. It's formatted at JSON and we get things like the wiki data description right there and we get the lead image for the article and for the lead image we get URLs to three different resolutions of the image so that the client can just decide which one to use based on the display size or bandwidth considerations. If the article has a recorded pronunciation of the title we can get the URL of the the audio file for that. If the article has a geo location tag we get the latitude and longitude and that's all followed by the actual HTML of the sections and another thing we're still working on is including a full collection of all the images and media in the article not just the lead image with similar multi resolution URLs and captions coming from commons things like that. So again these are all things that would either require additional network requests or some complicated dom traversals that no longer have to be done by the client. So the Android app is really serving as a proof of concept for what it would be like for an application to get its content exclusively from this kind of service and the results are pretty encouraging since we only need to make a single network request to get the content and all the metadata our time to first paint really just comes down to the network latency for fetching that single response and because we get all that metadata along with it we can have native components that consume the metadata and behaviorally smoothly and responsibly and by the way when you tap on links you get link previews like this which we've had these for a while now but these are you know now hooked up to the content service as well and we get a short extract from the article and image gallery of thumbnails that you can scroll through so anyway in the coming weeks we'll be rolling this out to a small percentage of our beta users and then ratchet up from there and see how the service performs under increasing load when I say you roll this out I mean roll out the version of the app that talks to the content service the app's been around for a long time now and also the content service I'm running this from my local machine right now but it is already deployed to production I'm just running the most bleeding edge version over here so yeah I'm looking forward to those results and one other thing I wanted to highlight from the app is the integration of the new maps functionality so previously the app had a nearby feature where it showed you a list of articles that were nearby you well now you get an actual map with pinpoints to articles that are nearby you and of course this is using our new open street map tile server and that's working pretty nicely you can tap on any of the markers and you get the same kind of link preview that you see one tapping on regular links you can pan and zoom as you would expect and it doesn't necessarily you don't have to see just articles nearby you you can you know zoom out as far as you want and go to a different part of the country or the world and zoom in there and get articles near there so we'll be rolling this out to beta in a week or two at the most followed very shortly by the full production rollout so yeah I think I ought to do for now so it's and still doesn't have the Wikipedia app installed I would highly recommend it and not at all biased towards the app right thank you let me ask a couple questions that came in on our IRC the first one what is the taxonomy based on are the metadata are there metadata for the schema I believe that's based on the JSON you were showing earlier so the taxonomy of the JSON is just something that we basically invented based on the things that we needed at the time it's been evolving you know for a while now and yeah we've we've gotten feedback not just from our team but from the iOS team and we've been asking the mobile web teams also to sort of comment on what they might need in the future and how the contents that were returned by the service can evolve even further cool and a question from us our hovercards and pop-ups are the extensions going to use this they certainly can I don't know what their plans are for this but yeah like one of the goals of the service is to be consumed not just by the Android app but anyone who wants to consume it it's it's much better than the Media Wiki API and has had another question which was you are you parsing the content the HTML content for the pronunciation or is that storage stored somewhere so if you're talking about the audio pronunciation URL so that's actually comes from the HTML because that's encoded somewhere into into the wiki text the content service runs on top of Parsoid so it takes the output from Parsoid which like takes wiki text and transforms it into HTML with some really nice classes on top of it and that's what we feed into like a temporary DOM and parse things out of there so that's how we get those URLs minute thank you Demetri this is fascinating I'm curious how easy it is to go from these new developments with your geo tags into something like Street View or Google Maps or Google Earth and even ahead into a kind of 3D virtual world potential from Wikipedia interesting that's a good question I mean in the Android app we could easily do something like you know if you tap on a button it would bounce you out to the Google Maps application or Google Earth with Street View with that location fed into it so you automatically go to say the Google Earth app and get that Street View is that's sort of what you meant all right and I have a question the nearby map feature looks amazing at a timeline for iOS release at all I can't speak to the timeline for iOS you have to talk to the product owner for that that would be Josh Miner very well thank you okay thank you very much Demetri let's give him a hand well thank you so our next lightning talk will be by Andrew Green on banner history hi thank you so much I just um by the way I just uploaded the slides to the media wiki to the media wiki page so if anyone would like to refresh that I have this um it's like a just-in-time slide creation system so it's there right now but can you see this those who are also looking on the hangout the slide there you just see you are you able to share your screen so oh I thought I was okay one sec okay how's that we go we see it thank you perfect okay yeah fantastic thank you so yeah I'm gonna tell you a bit about the banner history logging system the new banner history logging system in central notice which records banner displays a history of banners displayed to a user in a way that we think respects user privacy so just a quick overview of the initial requirement which was more data on banner history on banner display history basically users actions in relation to a banner clicking on donate or on any other action requested by the banner or closing the banner depends on more than just the the content it depends on what banners the user has seen before or what banners they've not seen however previous systems I didn't really get very much data about this in some cases there was a count of previous banners seen depending but there was no information on the timeline so they could have been seen a long time ago or recently so more data on this has the potential to let us target banners much better and thereby show people less banners here's some really specific questions that the fundraising analytics team was looking into when they asked for this feature one was to find out how effective the actual the current large banner first approach is I guess most of you probably know that for fundraising a large banner is shown especially or only to users who have never seen a large banner before and then after that is a series of other kinds of banners that are shown there's a really important question here how does the cost of continuing to show banners or when does that the cost of that in terms of annoyance in terms of disruption of user activity when stat outweigh the possible benefit of another donation a more general way of framing that is when does banner fatigue set in when are you just too tired of seeing banners especially fundraising but maybe banners of various types so they just have no impact on you anymore obviously that point we want to stop showing banners and there's really a huge possible array of the huge huge amount of practically infinite possibilities we could think of for how to target banners you know time of day wiki lots of different logic we could put into the to run on client side without invading user user privacy so this system of logging and viewing the history of banners seen is basically a first step to explore the universe of possibilities here and and eventually have more effective banners and thereby show less banners in the end so here is an overview of the banner history system as it is now so when you enable the system for a campaign the logger will start accruing log entries in the user's browser in local storage so with this the user can see everything that we store and knows exactly what we're storing and what we're not so there's no tracking of users via unique ID it's all local and visible and here are the main data items stored which campaign sorry which campaign the user is on which banners being shown the approximate time and date of the display of the banner and this is limited in accuracy specifically so that there's no way to correlate this to server logs so there's no way to find out from this log you know which page the user was viewing or which user this is and it's a logged in user which is actually not relevant for fundraising because only on almost users see see banners for fundraising but in any case for any kind of community or fundraising campaign there should be no way to correlate the history accrued here to the user activity on the weekend anyway and another data point that stored is just the status of the banner in terms of being hidden or shown because of this logic whereby you know sometimes the user is selected as being in a campaign but there's logic that runs it said okay it says okay let's not show the banner now or let's show it you know two or three banners over a set of page views and then let's stop showing them for a certain amount of time so data about what the result of this logic was on this particular page view is also stored in the banner history log data not stored no information is stored about article scene or user action taken nothing that can uniquely identify the user no generated unique IDs either the details of how the data sent back a percentage of users currently it's at 1% for the campaigns where this is enabled so this percentage is is selected randomly and for those selected users the full history or as much of the history as we can cram into an event logging call is sent back to our servers then also if a user clicks on donate on a fundraising banner their banner history is always sent back and here was a tricky thing that we were working that we thought a lot about tried to get relevant data while it's still protecting user privacy when a user clicks on donate we had to generate a temporary unique ID that is passed on into the next the systems after the the banner for collecting donations and that allows us to correlate given history to donations however this unique ID is completely temporary only stored in memory it's never persisted anywhere on the client so it can't be used for tracking the user subsequently and even though the user is subject subject to a different privacy policy once they click to donate there is still no way that we can correlate this banner history to any data related to their wiki activity and finally just to mention that this this new feature was developed in in tandem with other changes in central notice specifically a system that allows us to run code before a banner is chosen this was important for other performance improvements that we need to do in central notice especially the last one which is upcoming which is the removal of the dreaded special record impression calls which are calls that that are made to record with without sampling in the case of fundraising when a banner is shown but these are these are this is the old system of recording banner displays which doesn't have any history associated with it so this is going to be removed and the new system of running campaign specific code before the banner is shown helped us do the banner history logging and a few other things including the removal of special record impression which is pending but coming up soon and also we've had to make some changes in the algorithm for choosing a banner and there's also a bigger factor of client side central notice code so this is all kind of wrapped up together and planned as part of a single roadmap that allowed us to get the feature out and you know bundling these important changes that were not really very much user-facing except in the sense of improved performance but that were quite important nonetheless and that's all thank you very much right thank you Andrew we have a question IRC how many browsers support local storage yeah that's a great question so of the browsers that we actually support for showing banners initial an initial check of this is actually it's a very small number that don't support local storage it's point one percent this data is obtained by the fact that when the when the banner histories are sent back for users that didn't have local storage enabled if the if the user was in the sample to be sent back then instead of a banner history we just get a message saying this user didn't have a local storage enabled so that the idea is that will allow us to take into account the percentage of users that don't have a local storage enabled when we you know when we do statistics on the on the banner history data and so an initial sample of a one-hour period with about 14,000 banner histories sent back I believe I think the actual percentage of banner histories that said no local storage available was point zero six percent however that those are page views not actual users so for example if a user doesn't have local storage enabled but went back several times to the site and just happened to fall into the sample several times then you know then then they may be sending that message several times to to the log so so it's it's hard at this point still to correlate page views and banner history sent back to actual users I hope that with some statistical processing will be able to get better correlation there but in any case it's a very small number of users that don't have local storage there are browsers that such as opera mini that people use and that don't should don't don't support local storage but at least most of the time those browsers don't show banners that either and anyway the case is the same for old Internet Explorer browsers thank you any more questions all right thank you Andrew yeah my pleasure thank you so much so our next speaker is local Danny and he'll talk about the community tech survey hello there it is hi I'm Danny horn I am product manager for the community tech team here at the media foundation you're not familiar community tech we are a fairly new team just got started in July and we are here to meet the needs of the super active with media contributors find features and fixes that's going to help them to be productive and so we just did a week long tech support satisfaction survey sent out to a thousand people and the idea was for us to get a baseline on how satisfied contributors feel for the level of technical support they're getting from the foundation and then also to figure out some of the areas that we need to figure out and look more into so super fast through this this held in October we look at 10 with media sites who is nine Wikipedia languages plus commons we invited a hundred people randomized from the most active in the last 30 days and then they took a survey that was on a Qualtrics you we actually got a nice response to it a thousand invites sent out we got 467 people responding Russian for some reason like immediately we got a whole bunch of stuff from Russian Japanese Portuguese and French did not like our survey and so here's just the super big number one question was just overall how satisfied are you with the technical support you receive there's some variation between the projects the big number that we're looking at is 48 percent saying that they are satisfied or very satisfied which is not an awesome number but that's we wanted a baseline now we have one so then breaking that down a bit we asked people how satisfied they were so from a scale from very satisfied to very dissatisfied about some of the things that the foundation could be helping with and so you can see here if you look at the green and the purple together these are much much better than the other so the stuff that that we did well was reviewing recent changes helping with contributing new articles and managing the stuff that people really said that we were not performing on was helping to identify content problems supporting wiki projects and then supporting gadgets and bots and third-party system now by project this was also super interesting again the green and the purple show either satisfied or very satisfied Russian it just loves it's amazing 60 percent I have no explanation for this by the way Russian loves this with 60 percent and then there's sort of a four-way tie for number two English French German and Portuguese but again it's like it's 51% satisfaction so we're not crazy about it the the projects that reported the lowest satisfaction there was Japanese with just 26% wikimedia commons with 38 and then Chinese over here 44 definitely there's a qualitative piece that I'll talk about in a second but we definitely saw dissatisfaction from comments both you know here in the ratings then also as you'll see with the quality of answers Japanese just because I find this interesting more than any other project tended to vote for stuff in the moderately area I found that striking we're wondering about sort of cultural differences where in Japan possibly you sort of discouraged from going to the extremes so that might sort of relate to why we see so few in the extremes and so much in the middle but that is just a random guess for me and then this seems super interesting to me so this is how satisfied people are with typical support by how many years they've been active and as you can see the sort of one year or less where it's kind of 50-50 and then as people have been involved with projects we see less and less satisfaction there's a couple interpretations that we can make out of this one is that just the more that people get to know the project more that people get to know the wikimedia foundation they like us less I don't think that's necessarily actually the case another interpretation is that folks who were here you know seven to ten years ago that actually predated the foundation really existing and providing support we'll see some of that in the comments as well that I think there's some kind of generational difference where folks who have come in and joined the project when the foundation exists and is providing some support start to see it differently so yeah the quality of stuff what I have right now it's we're getting volunteers to translate all the all of the different languages right now I've got English I've got comments and I've got Arabic just to give you a little taste of the kinds of things that we're seeing one thing that's super interesting on both English and comments are these questions about not just what is the foundation doing but also what is the foundation's role in working with other projects so here's an example of somebody asking about it would be great to actually have a direct line to the foundation a few people said this that I'm not necessarily that technically savvy and I have no idea where to go to ask for help so that's something that beauty tech is definitely gonna be working on and then there's also some folks who say really the foundation should only stop in when there's a problem that needs to get fixed the fact that I haven't seen you kind of enter the wiki much is fine because if it's necessary that means there's a serious issue but folks are still asking for ways to help you editors and then also there's several requests for different kinds of gadgets and bots for work on common so remember I said we got very low scores from Commons these are some of the reasons why comments definitely some of the people there had really strong opinions about the foundation should stay out of our wiki so the WNF staff you're there to make servers stay on aren't these things handled by the community rather than foundation so this is just a small sample but I think we have a troubled relationship with Wikimedia Commons and I think that's something interesting that we ought to be thinking about and working with but there are also folks who do had some specific requests and so this one is about identifying copy bio vandalism faster we got a couple of those for the Arabic comments it actually was really really different didn't see anybody sort of self-reflecting on the role of the foundation they just really want help and they want help with a whole bunch of things so some of it is software related so they want help with content translation which we're starting to do as a foundation we need to deal with how hard it is to edit articles and then some of it was just about kind of community problems where there were a few comments where they said the administrators are pushing new people away it's very difficult so it's interesting to kind of see that coming from a wiki that I think the foundation hasn't necessarily had that much interaction yet so the thing that we are doing next is we've got a wishlist survey that is coming up we're actually we're through the proposals phase and we're now about to get into the voting phase where people are actually coming and saying here's the technical support that we want and they're voting on those projects one thing that's interesting is we're seeing some super different things there compared to the kind of comments that we got on the other survey watch lists are coming up a lot we really want to see changes to the watch list and then another nice thing that's happening with the wishlist survey right now is that there's a very strong interest in specifically making a couple features for comments so we might be able to kind of work through some of the issues that the foundation has had on comments so those are questions about this things will be not on IRC yet any questions here thank you for the presentation always felt that it's it's really hard to ask or major satisfaction without being specific about the things that you ask for if you if you ask people to evaluate the local government it's very hard to answer because they have some good programs and they have some bad programs and they have something in between right did you actually make sure that your respondents were well informed about the services that the foundation provides or make the questions more specific about how how do you evaluate or how much are you satisfied with certain things breaking down by by categories yeah there were three big questions that we asked the first one was just a general overall satisfaction where we're looking for sort of seeing this and then second one was these was specifically asking in the following was this eight nine areas how satisfied are you with that and then the third question was sort of apart from that what else do you think we should be helping with so it was kind of a range of thank you I have a general question so what's next right there will be the wishlist but how will you go about prioritizing and and what is your bandwidth for all all the requests that you'll get oh we're fixing everything no we we're totally on it we got please we got like a five person team it's gonna be amazing now the wishlist survey is is gonna give us a prioritized backlog to start to respond to and work on we'll have to sort of evaluate the stuff at the top and then that backlog also is gonna be shared not just with the community tech team but also with developer relations and community engagement teams so that we can encourage volunteers to come and develop some of those also some of the data that we get may be interesting to the other product teams at the foundation so there's a lot of work coming up but it'll be really nice to have like actually a backlog of stuff that we know the people really want and when will be the next survey oh and so we'll probably do another tech satisfaction survey like this in a year after you've had some time to do some stuff and then we'll see we'll see what that looks like right thank you thank you Danny hand cool so our next speaker is Scott who's gonna give us an update on a donation to wiki data hello everyone I'm Scott McLeod I'm the founder and developer of World University in school which is like Wikipedia in seven in 300 languages and MIT OpenCourseWare in seven languages and Yale Open Yale courses in one language it's all under the umbrella of creative commons licensing so this is where Wikipedia begin this is where World University in school began and of course I was teaching on Harvard's virtual island about seven or eight years ago we're like Wikipedia with best STEM centric OpenCourseWare we want to offer online accrediting degrees in all countries main languages so we donated with with World University in school to wiki data in for its third birthday on October 29 and what is it well the first perspective is all languages we would like to be in all 8,000 languages as wiki schools for open teaching and learning you could teach to your web camera add something anthropologists linguists could go out in the field to a language where there are only 10 people left and in Google Hangouts distributed people collect that language aggregated added to machine learning we would like to be in all nation states for accrediting university degrees wikipedia I think this 259 countries by population and UN has 160 170 180 countries we would like to have these are all the sort of main sections areas there are about 10 main areas at World University in school think of them as a template as a design as a plan as we begin to develop in wiki data you at World University in school and admissions registering at World University in school potentially five six billion people if we're all teaching to our web cameras we're also planning again bachelor PhD law medicine it's not up there I be international baccalaureate high school diplomas we seek to matriculate our first students in English then in UN languages and large languages so students would be studying from their high school rooms 50 from Brazil in English 50 from India it perhaps in beginning in 2017 for an undergraduate degree that's MIT open-course horse and subjects add a subject you have an interest in green chemistry add something about that you have an interest about a certain species of birds add that as a subject page you have a creative subject in music create that and create add MIT open coursework courses at Yale open Yale courses hop into Google Hangout courses in schools potentially aggregating also great universities open courseware there about 50 great universities already listed on this page everything from Oxbridge to universities in France to universities in China they have all open educational resources aggregate that begin to process that with machine learning over time as well so we become a kind of meta resource for all great universities online as wiki a huge vision a music school at World University in school so we would like to have all instruments each a wiki subject page or a school in all potential languages nearly 8,000 languages so from each wiki school or subject one could add the greatest YouTube's one could interactively play music in something like a Google Hangout could you do this right now even in real real time since there's lag in Google Hangouts from cell phones with the video in Google Hangouts probably library resources we would like to be in all 8,000 languages 7,930 language 938 languages per glottalog right now each becomes a school we would like to aggregate all the databases of libraries to each language potentially integrating Europeana DPLA all under the CC license we're looking for open and free resources maybe we'll network with state libraries with their somewhat public mission at least to their nation states museums all 8,000 languages potentially each ad databases ideally this will emerge in a real virtual earth so we'll be able to go into Google Street view Google Maps Google Earth open light open simulator second life and pull a book off a shelf and look at it or in other ways such as on screen so this becomes potentially interactive and 3D hardware resources possibilities add all interesting creative open resources such as Raspberry Pi or someone in Africa who what was his name the MIT person Kelvin Doe who was brought over from Sierra Leone who is creating interesting electronics out of a dump he probably would add interesting resources in a YouTube for example add these from every country and its main languages educational software add all our creative projects to each languages wiki subject page the wiki subject pages in each language will grow into hundreds thousands of pages as schools grow in each language anticipate and plan for machine learning and interlingually especially it's the interlingual aspect that world university in school potentially is mind expanding around so wiki data the new database the three years old wikimedia asset is in 300 languages it's interlingual it's anticipating machine learning sparkle and owl research add a research project you're interesting integrate all the interesting bioinformatics databases out there if you need clinical trial if you're a researcher at UC Berkeley or at MIT and you have a rare disease with people distributed across the world send them packets in the mail and create set of protocols that you might even monitor over group video so you at world University in school with all the basic research fundamental research can emerge in each country and in each language potentially and again anticipate and plan for machine learning and interlingually as well we have a foundation we'd like to be a growth story minutes for wiki data for wikipedia for Google for the web so the all languages thing and the all countries thing opens up all kinds of horizons for innovation that will expand potentially the web in very far-reaching ways we plan to a credit in 200 countries and we would like to become the Harvard MIT Stanford Oxbridge of the internet so here are some questions I thought might be relevant to this talk they're on the slides on the lightning talk page and are there questions here now because the interactivity is the questions are most helpful for creating knowledge-generative conversation yeah I got a question from IRC which is have you looked into improving wiki university org like using that as a platform for courses great question our pedagogical model is on the degree side not the wiki school side of 8,000 languages is for the undergraduate degree basically a read college pattern required first-year class required junior qualifying exam senior thesis on the MIT OpenCourseWare which is in seven languages the other languages besides English the six languages have maybe 20 to 100 courses and they're typically translated by universities in those countries so we would we have aggregated we've wiki universities at world university in school but it's integrating the plans the visions the templates that's something we would do more fulsomely with wiki university I did come into conversation three years ago with someone we have a very different pedagogical model who is from wiki university and we didn't pursue that well any local questions not yet so but I'm also curious right like how are you working with wiki data right now like you donated knowledge to them right have they ingested it so wiki data in the past year or so has been attempting to focus its external projects its sister's projects and they've developed a number of different ways we're out of the box in some ways so we donated the 720 wiki pages that's that are in wiki wiki right now mostly in English besides the MIT OpenCourseWare in six other languages to them and we seek to develop in wiki data having developed that and with a media wiki as a front end and how we do this and how we grow from being from within wiki data now which is cc and open with this huge template plan this huge design plan this huge growth story for the web is a question that's unfolding cool any other questions no people are light on questions today cool well thank you scott thank you very much wikimedia foundation all right so this concludes this month's version of lightning talks thank you very much