 Om dat eigenlijk online te houden is het eigenlijk een hulp. Dus we hebben het online afgemaakt en we hebben het in de archijf geplaatst. Dat is waar je kunt monitoren, waar je het gewoon afgemaakt kan houden. En dan zijn de mensen er niet blij mee. En dan hebben we het weer online geplaatst. Dus we hebben een HTML-snapshot opgemaakt en we hebben de files online gehouden. Dus er zijn geen veiligheden, maar de site werkt niet meer. Het heeft wat griezen en wat buks. En voor de tijd is het goed. Zodat het browser begint te veranderen. Ik weet niet of je nu een website hebt gevisiteerd van, laten we zeggen, 95. Ik weet niet hoe het gaat renderen. Ik was eigenlijk wel heel blij dat het zo mooi kwam op mijn kroom. Maar ik denk dat dit in 10 jaar niet de keuze zou zijn. Ik denk dat als je dit op de tablet kijkt, dan zou je dingen vervaren. Dus wat is m'n punt met dit? We willen niet veel over dat website. We kunnen gewoon wat snapshots hebben over hoe het looked. Maar dat hele load van informatie, dat is op dit website, we willen het doen. Je moet het in de archive houden, want dit is een historie van Gent. Zijn met jouw organisaties. Ik weet niet hoe jouw organisaties hun oude porties, de oude facts die ze publisheren. We hebben grote archives en de hele fysieke partij, het is een grote traditie van, oh ja, ik hou deze papieren veilig en dan op dit niveau. En er is een plan om het te migraten en documentmanagement enzovoort. En ook wel, we hebben sites die gewoon veel data in funky databases hebben geplaatst die niet werken in 10, 20, 15, 30, 100 jaar. En als we dat data linken, gewoon de pure informatie terug naar links open data, dan kunnen we uiteindelijk hebben een format waar de format is anders dan de technologie. Het is een format en het is geen technologie, dus je kan het uitvoeren van verschillende technologen over de verschillende jaren dat dit heel inspiratie was. Dus terug naar de Gentse fysie 2017. Dus we publisheren alle data als links open data. Again, I want to reiterate, this was not a technical problem. Well, of course we did that. Oh, there were lots of technical problems involved. But the main work here was also like putting all the taxonomy, organizing our data, organizing the data from a data architecture point of view versus like, okay, we need something that outputs our data. So now if you could, there's like still, there's data being published as JSON, but it's like linked semantically. So you can sort of get some semantic meaning of it. And it's fully linked as linked open data, from our linked open data platform. So what did we do? Then two weeks before the Gentse fysie started, I have like this little lab and I work with students there. En they said like, I said like, we need to do something with the linked open data against fysie because it's cool. And then they said like, okay, we can make a chat bot. So basically this works as a sort of stateless bot. It runs on dotnet core. So basically you run it on Linux in a docker if you want to. And what does the bot do? It talks to you. Well, you ask him. You can give him questions like, it says like, okay, choose one of the squares or you can choose your location on a map. And then if you say like on a map, you get like these buttons. It does not really talk. It has no natural language. I will come back to that. I will come back to that. So it has like these big buttons like what's to do and it generates what's coming up next on different squares. So this is nice. You can also get like more information and then it gives her that answer. So how does this work architecturally? Our bot does not store any data. It just relays data back for the person who asks it. Then it asks our triple store for like, okay, give me some results. And then it packages the triple stores back. I need more time for this. Five minutes, okay. So then it just, you ask him a question. He asks a triple store. He says like, oh, the triple store is set like this. And okay, and that's how you need more. There's no data in the bot. You can scale it horizontally if your triple store is stopped or not. We had a few thousand visitors on that. We had some interesting reactions. Because if you say chatbot, people tend to go chat with the bot. En we had, we learned it like four languages, like the main languages. And we put like, like, like Gentian also in there. So we learned a bit about how people would react with a bot like that. And they tend to go talk. And then they don't, also they give compliments. Like we're doing a good job. Thanks. I'll pass it on. So we published the bot at like everything, which is built on Lab9K. We published it at GitHub. So you can try or experiment with it yourself. So it uses buttons and no natural language. Why is that? Basically because I'm old. And in my day chatbot, you type with them on IRC. I don't talk with my computer. But it doesn't do natural language. En we had to build it in nine days. So if you then want to learn how to do natural language, I think there's some companies out of wealth here. I don't think they would put it up in nine days. That's fine. But I want to come back to the linked open data thing. Because these big guys, they are developing the natural language both. En let's hope that some small guys also jump into the game and we get some nice interesting competition between the big part and the small part. And these guys, they won't visit your open data portal. They won't visit your workshop on how you open data. They need to have linked data so their bots and their AI can work and spider and derive meaning from all the data sets you have. En if we produce the internet with JSON files and XML files, this will help, I'm sure. But if you want to accelerate this and we want to plug in to these systems instead of like the small web builder systems, the small specific app builders, then you need to have linked open data. For us, the Gantse feesten, we get 1.2 million visitors every year. App builders are still interested in getting something in the app store so they can have some nice refutes and very high-rated, much downloaded app in their portfolio. En that's because it's the Gantse feesten. They want to do it for the Lichtfestival, which also gets like 1 million people. They want to do it for several other big events in Gantse. En if we want to come to a world where you can say like Siri, I want to go to the... I want to see some hip-hop at the Gantse feesten tonight. That Siri will notice based upon the linked open data and it can tell you like, okay, this is going on, this is going on. And because your data still has this unique point, Siri can tell you that the thing is gone or it was delayed or whatever. So I think we need these bigger ones or very small, very well-motivated teams to build like that next generation of interaction en people talking with the bot. I don't think a small app builders can do it still. So there's something else. You can save it. This is also like... Okay, I'm going to do demo things now. Is the mouse supposed to be... There we go. Oh, that's all right. But that's... I'm very curious on the video now. It's going to be the best video of the event. So we built this. It's like a small demo. Also if you go to our GitHub page, you will see how to communicate with linked open data from different languages. This is a basic JavaScript front-end. So it has like... It's a host on GitHub. It has no real server-side things to it. You can select a date on it. And then you can hit the search button. And then the results pop off. So this thing is just formulating your request. Then rewriting it to a linked open data request. It sends this linked open data request to our linked open data server. And then passes the results back in JavaScript to you. So you can actually build like this full website met a search and everything on linked open data in purely JavaScript. So these are already opportunities you have today. If you go to... I don't know if it's linked, but if you look at... look for this, it's probably not linked. If you look for lab 9k, it's a little project you're working on within DigiPolis. All these codes are free and open. So you can go to the GitHub en you can see it all. There's like a whole folder of querying linked open data in different formats. And you will also see this small demo there. So, it's good. Bye, thanks.