 So guys, good evening. I will try to do a five minutes talk about reasoning on the web. So basically, today I'm going to talk about bringing reason to the web using example of the project I'm working on now. I come from Brazil. I'm doing an internship here at iPall. iPall is an image and services lab. It's a French-Singaporean lab that works with assistive technologies using reasoning to the web, also machine learning and other kind of intelligent approach. But first, before talking about reasoning, let me talk about what we are doing in my project. We are trying to propose something for ambient assistive living. What is exactly with ambient assistive living? Basically, try to think about your grandmother living alone in her home. We are proposing some assistive technologies to support her independently as long as possible inside home. And once she needs to go to the retirement house or nursing homes, we provide some services and some technologies to improve their lives there. So what exactly we propose to ambient assistive living? A smart houses. What are the three powers of our project? Same thing. Basically, we have a lot of sensors at home and also a lot of sensors at the room and the retirement houses. That sounds like motion sensors, bed sensors, and bed sensors. We put everything this data. We set all the data using house purifies. So we have a house purifier for each house and also for each retirement room. We set all this data for our server. We have a cloud-baser server. Why? Because we want to have a lot of houses being monitored for our system. In our server, we have a reasoning approach that's trying to identify the activities of the users based on events that you sense from sensors and provide some services based on these activities. Let me give you an example. If your grandmother is cooking for a long time, maybe it means that something is wrong. You should not find a family or you should not find a doctor. Or try to think if we did find that your grandmother falls down on the floor. We should call the doctors. We should call the caregiver. We should call also the family. Let me talk about reasoning. We have a web-based system, but our focus is basically research and reasoning. What is exactly reasoning? The idea is to, given a bench of events, a sequence of events, we should identify what kind of thing we should do in response of these events. Like my grandmother falls down on the floor. We identify that based on the shape of the body. And then I propose to call her to notify the caregivers. When we do this, we have a knowledge-based driven reasoning using Semantic web. So we are on the web. We use the representation of data of this Semantic web and a knowledge-based driven reasoning. We use the notation tree. It's a notation for Semantic web. For inferences, for identified activities of the user, we use ORI reservoir. It's basic. It's a rule-based one. And also we have ontological representation of the environment. So we have a knowledge representation of the house and of the user. What is exactly its ontology? This name is really crazy. Basically, its ontology is just a graph of things. Let me think about internet of things. We are doing, we are coding, Semantic web, concepts, and internet of things, and providing services for our people. Basically, ontology is a graph of things. It defined it for triples. What exactly triples? It's how we represent data in a Semantic web. Basically, triples is a thing. How I can represent a thing. We have a subject, a predicate, and an object. Based on this kind of concept, we can say that a predicate and a triple is a thing. So we have a lot of things and representation of things in our ontology. Let me show you one example of this. This is really meaningful. Here, its ontology is how we represent data in our system, how we represent data in Semantic web. It's important to say that each thing in Semantic web has a URI. Each predicate, each object has your own URI. So this data is open. So you can take this data as well. Here, I can show you the URIs. So we have this representation. And here, the representation of the triple. So I can say a house is a class. A house is a subclass of environment. A house one is a house. So kitchen is a type of kitchen. And this instance of kitchen is part of the house. So we represent, we design our whole data using triples, saying this is this and this. So it's using subject, predicate, and object. And this is an example of rules. I told you that we haven't raised any basal rules. So we have the data and also we need the rules. Basically, what's the rule? We define one sequence of event that happens. And also, you define a consequence of these events. For example, you identify that your grandfather, it's on the bathroom for 30 minutes. And also, it's less than midnight. So we have two events. And two events, you have a consequence of that. What are the concepts? Maybe notify the family, notify the doctors. But we have to bring this for the web. Basically, we have two packages that we need to install. iServer and triple store and tribo.js. All of them were made for Cuban verb off. It's a Belgian research in Semantic Web. Basically, i, it's a desktop-based reasoner. But you can bring it for the internet. Basically, you use an iServer. It's an NPN package for Node.js. So you can bring it. And also, you can bring the concept of Semantic Web using a triple store. How to access the triple store, how to handle this data using Node.entry.js. SID idea is to discuss about research. Obviously, we researchers, we love schemas. We love to draw architectures and generic ideas. So this is some approach. Basically, our reasoning approach to the infactivity of the user is quite heavy. We have a lot of steps. But I bring to you something that it's a goal, that you can put in your application if you want. Do you want an intelligence in your application? Do you want a rule-based reasoning in your application? Do you think that it's important for you to infer data based on data that you have? OK, we can try to do this using this schema. We have inputs. Basically, you have input handler to handler. The input that you have, you have the cart text. Cart text is the main idea of your reason. If you get this data and put this data in a triple store, triple store, basically it's the store of the data that we have. Once you get this data from a triple store, you can set it to the reasoning. What my reasoning means to have, to infer data, to understand, to predict a base on the rules, you just need to get the rules that you already designed and the data that you have and set it to the reasoning. And what's your reasoning you'll do? Basically, you infer data. My grandmother is in the kitchen for 15 minutes and the television is on. So this is your data. Where's the rule? If my mother is in the kitchen more than 10 minutes and the vision is on, please call the caregiver because probably we have a critical situation and we have a warning situation. And this is the model that we should follow if you want to implement to bring this intelligence for your application. And then tutorials. If you're really interested to go through that, take a look in the tutorials. All of them is based on a semantic web and reasoning and also made for the Belgian guide to view the researcher. Thank you very much. I'm Joao. If you have some thoughts about Reasoner, I just tried to give you a generic idea of that, but I have a lot of things to do and a lot of things to discuss. And thank you very much. Then the age law, right? So are you doing processing and everything on the age law? So are you pushing out the data? Do you have a lot of processing that you're doing? Let me show you the last one. Toucher here, here, here, here, here. This one is cool. So basically, we have two sides in our system. The house side, there's the sensors that we have, okay? We centralize the sending of sensors in our hasperia pie. So the information of the motion of the doors, sensors of the bed sensor, it centralizes it in our hasperia pie. And we send this to the cloud. And in the cloud, it centralizes it. We have a centralized reasoning approach. So we get this data, we process this data, we put it in the semantic web, we do stuff using triple storage. And then we send this to our reasoning and get some results, like a notification. So you're having a lot of data that's captured on this. Yes, a lot of it. We have a problem with this latency of data because there's an e-group, all right. Not yet. What do you mean? Yeah, now our first prototype, we have a nursing home in France with 20 rooms. And until now, we didn't have any problems with some bench of data of the quantity of data. But we are concerning about it because our next prototype will be 20 homes and 20 rooms in the house. So we should think about how to handle the bench of data. Actually, we didn't think about it yet, but it's a good thing. Because you know, that hasperia pie is a feature for something that you have in your house. Yeah, but we have one transfer for our house. So it's pretty okay. We didn't have any problems with that. Maybe the problem is seeing all of the data to the parallel and process energy. That's one more quick question. How exactly do you determine the rules? Isn't it based on like location or does it learn and change for them? This is a quick question. Now, in our first prototype, we are not learning about the user yet. What we do? We design some rules for basically, okay? We run some experiments in the nursing room, check with the nurses and the cardevers and the doctors if our rules exactly address the activity that they want. One example, for doing two days, we will monitor one user and then check if the activities that the user be corresponds to the activities that will be inferred. So actually our work is really closer to the nurses and really closer to the doctors because we need to validate our approach. But basically we designed the rules before put in system. But now we don't have any learning, but the next PhD students in the work environment. We will be back. And also we have a really good project coming that's supposed to be my research and activity is to bring the system also for the cars. So we will be trying to inferred activities of the user in the cars and provide services in the car. Okay, we will quickly move forward for the next, I mean, talks. Yeah. So any questions can probably do. Yeah, okay. They just finished with my email. So if someone wants to talk about it, I'm really interested. Drop in such email, of course. So this slide will be available while we share with you in the group. Okay, thank you very much. Thank you.