 Good morning, everyone. Welcome to Open Belgium. I want to quickly thank our three main sponsors, Mono Design, Microsoft and Agentschap, Binalands, Besture. But now I really want to give the floor to Freya and Lien. Thank you. Good morning, everyone. And welcome for our session. I'll quickly start sharing on my screen. All right, now we should... Yes, now we see my screen for our presentation about Probe. And Probe has a second title, Policies fit for humans and machines. So we can deduce from this that Probe is about policymaking and for policymaking. We first need the policymakers, which you can see here. These are policymakers from Gens, within the middle are a mayor and the president and vice-presidents of our city council. And then, on the other hand, we also have the policies themselves. As you can see here from the last city council, we have some policies themselves. But you can see that they're quite difficult to read. Nevertheless, they are important for everyone who is in Gens, which includes everyone who lives in Gens, who works in Gens, who visits Gens. And to get a more clear example or a more clear view on how these policies are important for the people who are in Gens, we've constructed three fictional examples of people who are interested in the policies to start this presentation. The first person is Sophie. So Sophie is 39 years old, works for a bank. And she also lives in Zwene-Arden, which is a part of the city that you can see here. And Sophie is a happy member of our community and she's interested in information about roadworks that are planned in her neighbourhoods because this influences her commute to work, this influences her daily life. So she wants to know more about roadworks. Then she's also interested in big construction sites in Zwene-Arden if a company wants to construct a new building or if anything will change about the view of Zwene-Arden the way it feels. And then lastly, she's also interested in events that are organised nearby because Sophie also really likes to party. I think you can all relate to that right now. Then our second fictional example is Mo. Mo is 53 years old and works for the city council, the people you saw in the first slide. And to prepare for a council meeting he needs information about previous policy decisions that have been made about certain subjects. He would also like to have information about similar policy decisions and about the history of the subject to be able to properly prepare for the city council meeting. And then lastly, we have Alex, 25 years old, volunteer for an organisation that's concerned about the environment. And Alex is very interested in funding opportunities to be able to start projects that are good for the environment, rules and regulations from the city of Ghent and policy notes and viewpoints from the policy makers. This is where the aldermen, the schepenum state what they want to do and the next coming years where they want to go, what they want to achieve which is very interesting for organisations. Now, if you remember from the first slide the way the calendar is, the way it looks like it's very difficult to get the information out of there for Sophie, Mo and Alex to get this information from the websites. So, this is what Probe wants to change. Yeah, so to tackle these challenges we started our project Probe. Probe stands for Proactive Open Barrett van Bastur proactively opening up the information in local decisions. This is a city of things project so it's funded by Vlaio and city of things programme and smart city applications so that's what we're trying to build here. Yeah, for real. Next slide, yeah. The main goal of Probe is to make the information of local decisions easier to find, easier to search and also easier to access and with that goal in mind we will be creating a new application and this application is both for internal stuff and as for external users. So, as an internal tool there was the use case of Mo. He wanted to prepare for his council meeting so that's a feature that Probe should support but at the same time we are also reaching out to the citizens of Ghent getting them more involved into policymaking and making them more familiar with what has been decided by our local council so Probe is really a project about being more transparent as a local government. The way we want to achieve this is by extracting as much structured information as possible from the decisions and the structured component plays a very important role in our project because the more structured our information is the easier it is for machines to understand it and the easier it becomes also to reuse in different kinds of applications. At the same time we know that lots of interesting information is actually residing into unstructured components. Often the entire context of a decision is published as an attachment to that decision so as a Word document or as a PDF and we also have for instance audio and video fragments of the local council so they are a real challenge because they are unstructured and with Probe we will try to explore how we can gain insights from these unstructured components as well. The two main technologies that we will be using in the project are linked open data and artificial intelligence. This is how the project was structured. It consists of five main pillars. The first one, the ASIS analysis we started a year ago in the middle of the first lockdown. The main goal there was to have a look how does the ASIS architecture look like what are the current applications involved in the publication process of local decisions how does the information flow what linked open data standards do we need to take into account. Then the next two they're really the core of the project. We have on the one hand a user research track which is a kind of major brainstorm that we've had with lots of potential users. We've interviewed many of them in a one-to-one interview and we'll also be sending out a survey. On the other hand we have a more technical work package so we'll be screening the wish list from the users and then we will try to translate this into a more technical roadmap and this roadmap then will be the driver for our implementation process. So these two both the user track and the reference architecture they're being approached in an agile manner so we've done a first round of interviews now we'll start creating a UI wireframe of our application we'll be going back to the users until we are satisfied with the blueprint of our application. Once we've finished the user research and reference architecture there will be a proof of concept by November this year and if that proof of concept is successful then we hope to implement a solution in 2022. Next slide. Okay, let me start by quickly guiding you through our current application landscape so you can see a screenshot here of our web application. The URL is at the bottom of my slide so abesleutforming.gen.be This is an online website it's open to everyone as a citizen you can easily consult what has been decided in the different councils meetings of the city. It also has a calendar view as you can see which meeting is taking place when you select a particular meeting then you can see the entire agenda and there is also a streaming functionality so if you want to hear how our politicians are debating about certain topics in their Lucius Gantt dialect then this is a must. Next. This is more in detail what kind of information is made public so on the left you can see the different types of councils some of the information is already available from 2013 other content has only been published quite recently making information about decisions publicly available can be done at three different levels and there is a Flemish decree that determines exactly what must be made public on our local website so either we publish only the titles or a summary of the decisions of the entire content. For some of the meetings we already publish the proposals even when they are in a draft status but others only get published when they are formally approved and as you can see we also have a quite broad coverage already of live streaming. The content that you see on our website is also published as linked open data it's published in RDFA that means that the contextual information and the semantics is embedded directly into DHTML and this RDFA is also harvested by the Flemish government because our way of publishing the information on our local website fits into a larger program LBLOD Local Decisions as Linked Open Data and that's a program run by the Flemish government and their aim is to build an ecosystem where all information from all local websites in Flemish is harvested LBLOD the program is built on standards and linked open data and the two standards that are used in the exchange of information towards LBLOD there are two standards of the Oslo program so the open standards for linked organizations program and in our case it's the application profile Publicassie Besleut and Mandat and Databunk so they really describe the core model in which decisions and meeting minutes and mandates should be expressed so finally we're talking about linked data so that means that we not only express our semantics in a formal way but we also try to link it up to other databases in the case of Ebesleut forming where possible we reuse the URI of the central database for mandates and also the URI of the Flemish codex yeah okay I'm publishing linked open data in the source application where the information resides so as we do here for Ebesleut forming this is totally in line with our overall linked data strategy that we have in Ghent one of the guiding principles there is that we try to adopt an interoperability first strategy both on the semantic level meaning you try to express the meaning of data preferably by using standard vocabularies but also on the technical level so we use really interoperable formats and both semantic and technical interoperability are two very important conditions if you want to integrate with different data systems so the fact that we publish already linked open data in the source application and not transform the data to linked open data in a process afterwards this adds up to the fact that we really want to encourage a kind of decent centralized approach of managing data so managing the semantics should be done by the data owner and our motto is to collect data only once and to reuse it wherever you can so by making the data interoperable you of course want to stimulate the use of your data you make sure that your data can also easily be reused and an example of this is of course a human friendly interface as we have right now okay so this was rather theoretical I'll try to demonstrate this with an example it's a pretty famous example for the people in the room who are from Ghent where are the sheep and no I'm not talking about the outer piece of Van Eyck but actually about real sheep the kind that is doing some grazing on green meadows they play an important role in Ghent in eco grazing and they're really famous because they have a sheep tracker and you can follow in real time the position of a herd of sheep that is walking around the city and the data are also available as open data on our open data portal by the way but there actually has been some local decision making about the sheep so if you look for sheep on our website you will find some decisions as you can see here yeah so here you see the meeting minutes of the city council it's not really clear but you can see here who was present at the meeting when the decision was published who voted in favor or against etc and you also see the link to the live stream this where are the sheep this has to be translated to a model and this is a more detailed view of the Oslo model I will not explain it into detail but you can see some core concepts here on the slide so which council meeting it is which items were on the agenda who voted in favor who voted against a description of what was decided and what was the motivation behind it for example yes next slide and this is then how the rdfa looks like so you see how the semantics in yellow it's indicated how the semantics is embedded into the html so it's added at the level of attributes so you can really see how the model is published on the html page okay so this was a rather quick introduction of the as is application I think it's now time to hear what our users had to say so the three fictional citizens that we presented at the first part of the presentation are actually based on results we got from our user research it's just the English term I constructed for Hebrew and so in our user research you want to discover what the needs are of the people who are interested in the policy decisions so the first goal of this user research was to determine like what do you need how can probe make the interaction with policy decisions easier how can we make it more straightforward but on the other hand also is there an added value that we can create through probe then a second and a third goal of this user research was first to gain support if you have the stakeholders if they're interested in the project and you have their support it's easier to get the results you actually want and to also construct a pool so once that we've been able to develop certain things and we want to do this in an agile way so we communicate back and forth with the actual users about the products that are developed in probe and this is an overview about the people we've actually interviewed so on the one hand there was a civil servant so the people actually worked for the city's administration that we've interviewed such as the residents of the city council the management support service who are most close with the policies participation communication departments and then more thematic departments and on the other hand we also interviewed some external stakeholders such as the press who is also very close with the policy making process members of the advisory boards some citizens someone who has a café etc and now what have we learned from this we've seen that knowledge about the decision making process is actually quite quite good especially for civil servants they know how a decision making works where they can find stuff however with the external stakeholders there's a lot of variability there are people who are more acquainted than the most civil servants with how the policy making process and then there are people who don't even know that there is a public website for instance and then we've seen that most people use the regular tools such as the epislate forming and also the regulations that are on the website of the city of Ghent but some of them also have some custom software or other websites where they find information then the main results we found like the main conclusions up to now because we haven't quite concluded everything is that a more intuitive search would be a great help that's something we've heard from almost everyone we've interviewed then a link between related policy decisions so when you look for something a certain policy decision you can easily find something that's related that's about the same subject there's also a need for receiving notifications for instance Sophie, if a road work is going to take place in the streets next to where she lives she'll get a notification about this there's also quite some interest in the spatial implications of policy decisions which location do they have an impact and to be able to look for themes such as mobility, environment etc so these are the things that we want to do the results of our research but what are the technical needs for this so I will try to translate this to some more technical components you can show them on the next slide so these are the technical components that we see but this is really work in progress, it's a first draft we've only started this work package in the beginning of the month so the next coming weeks and months will be devoted to exploring this more in detail maybe for the next open belgian session we will have more definitive schemes so this is a first draft of technical components so maybe the first one more structured data next slide so the first thing we'll need to look into is to define the scope of data that we will be using in our pilot probably we will be adding more structured data than the ones that we are obliged to make public on our website according to the Flemish decree content of meetings such as the Q&A and the committee meetings they are also available on our website we are not obliged to make them publicly available so we will evaluate how well they fit into the Oslo model so that will be probably the first thing that we will need to look into also we will need to determine how far back we go in history, will there be a cutoff in time but what is sure is that all of the structured data will be harvested into our knowledge graph and for this project this will be a virtuoso triple store next slide as I said before the challenge of the project will be to include the unstructured content we have for instance audio and video recording as you saw and within that recording component there is also functionality that allows to translate the speech even the dialect to text so this means that these recordings might be a very rich source to add to our architecture and another type of unstructured information are word documents or PDFs of course so you can see an example here so here are my sheep again this is a document on the sheep that are grazing in the grasslands of Bourgoyen-Ossimiersen so if you are able to extract the term Bourgoyen-Ossimiersen we might link it up to a location but we could also link it up to a theme such as ecology or eco-grazing and if all these concepts become interconnected we will allow for a much richer way of query next slide so these kind of things it requires that the information that is presented in a text form can also be processed so we will need to define in the project patterns to really understand the semantics the context of the text for further processing and then natural language processing is typical the area that might solve these kind of challenges so extracting terms from unstructured content trying to determine their meaning and according to some predefined categories classifying information such as time, names, location that is what we will do in the named entity extraction component but we will also be extracting metadata for instance from the PDFs or the words or text from PDFs so these are all done in the knowledge extraction part of the architecture next slide linking to other data sources via URIs might also be an important feature in our future architecture so in our current solution we already use whenever possible the URI of the central database for mandates and for the Flemish codecs but depending on the use cases for instance if we have user stories on extracting location concepts we could easily imagine that we will also try to link up to for instance Crop which is the base registry for addresses or to the generic platform for public domain information which contains information about road construction work so these things are still to be defined but it will be a very interesting feature as well then the project definitely will benefit from a good search engine I think one of the feedback elements in our user research was that the current search on eBus loadforming which is an exact word search was too limited so we will be exploring as well whether we can optimize or searching algorithms use for instance fuzzy search synonym search to help the user to find what he or she is looking for and then the last two components they have to do with the interfaces that we will be building on top of it it's clear that we want to make the data accessible in an open standardized and machine available manner so for our knowledge graph our Sparkle endpoint will be available and we will probably also look at the JSON API to make the data available for other applications and then last but not least we will also be implementing of course a user friendly UI on top of it what exactly will end up with that will be determined in the user research track it could for instance a bot with AI where our bot is capable to answer complex queries so what this will be defined the next coming weeks all right so as you've heard by now I already read the beginning of the project so we started defining what we want but the biggest part of work is to come so this is a high level overview of the timing of the next steps that we will take so in April we will launch the user research and we will communicate results on the website then from April till June we will explore the market together with the Agentschopp-Binlands-Bestuhr because they have similar interests as to extract information from policies and from September on the actual development will start and the project is launched in October 2022 so what can we do now if you are interested in policies from the City of Agentsch you can fill out our survey I will paste the link in the chat in a minute and then for me one of the most important things to mention as well is that we have to do this project we need a team of people who are very capable with very specific expertise the people from the team here from the city side that's from management support office and data and information and from the Stryknel Nehe team web and also it's an amazing team so if there are any questions we are here I see a question from Isabel I think you can start looking at it do you maybe want to can we give the microphone to Isabel I will also try to I can unlock the participants yeah okay cool because I did not entirely get the question Isabel is still there yes but everyone entered in listen mode only so they have to reconnect to the audio if it takes some time okay okay sorry it's typing yes hello I'm back sorry for this welcome back very interesting presentation and I was wondering if you maybe you did not because that's your roadmap at the end but is there any intention to apply that to the COVID measures as you know there's also kind of policies there's been quite a lot and as a citizen if you want to know which measure applies to you as a certain point in time it's almost impossible and so this kind of tool you are you've been describing there would be extremely useful if we could transform all these policies which are in different sizes or measures it is something you can put into a virtual assistant or a bot so that I can query a bot and say no I want to go to the restaurant today in that place can I go to the restaurant? I think it would be a very good application I think concerning restaurants I don't think restaurants are in the policies but I think it would I think it's a goal to make it possible that you know for instance with the face masks that it's easier to know where you have to wear one or what is the policy right now so I think no if you are interested in developing a use case around that in an international organisation called My Data with My Data for Pandemics where we look at how we actually want to use a virtual assistant with machine readable rules to actually support travel go to school, go to work whatever you may dream in the context of pandemics and I think your tool behind would be extremely useful so happy to have further discussions Yes I think that's a good idea it's quite difficult to understand sometimes there's just like waves in the sound of the microphone but yes I think it would be very good to continue that conversation somewhere because it sounds very interesting I'm not quite sure whether we can help it probably for that but maybe we can it will require for instance that we also extract concepts of location for instance you need to wear a face mask in the shopping pedestrian area then we need to be able to locate really what that is it's a very interesting use case I think for our future tool I'll send an email right away thank you any other questions or remarks I'll also paste the link for the survey here yeah intuitive search it's a broad term of course we're all thinking of a google-like search but things we want to explore is like of course synonym search fuzzy search so that takes into account proximity of terms and things like that so these are searching algorithms that we will be exploring into the project if there are no further questions I would suggest to maybe stop the recording and we can leave the session open for a few more minutes if people do come up with more questions we can still answer them but I will already stop the recording okay