 Check, check, hello. Hi everyone. Thanks for coming out today. It's my pleasure to introduce the next talk in the TSVP Visiting Scholars Program. This is Professor Christophe Claremont from the University of, sorry, my French pronunciation is awful, so don't cringe. Ott-et-metier, Institute of Technology in France. I have a bit of a truncated CV here. He is where he is a professor of computer science. He also works at the Naval Academy Research Institute and is also the deputy director of the IS Blue Interdisciplinary Graduate School for the Blue Planet, which whose goal is to push back the frontiers in marine science, marine technology and ocean innovation. So this seems like a big interdisciplinary group of marine scientists. He's previously worked as a senior lecturer in computing at the Nottingham Trent University, and as a senior scientist in geographic information sciences at the Swiss Federal Institute of Technology. His research is very multidisciplinary, but tends to focus on thinking about geographic questions quantitatively and using very big data to do so. So maybe some of what we're gonna learn today could be most aptly characterized as sort of like a macro sociology, highly quantitative approaches to understanding human movements using things like cell phone tracking, particle tracking, things like that. And he also has interests in entropy and trying to figure out new ways to describe entropy by taking space and time into account as well. So we've been chatting a lot about how we might use some of his approaches to characterize biodiversity, say in a spatial or temporal context. So with that today, he'll be talking about people, environments and sensors, cyber opportunities for exploring human mobilities. So Kristoff, please take it away and let's all give him a quick round of applause. Well, good afternoon to everyone and thanks Dave for your nice introduction, but somehow and really put my presentation in perspective. Before starting, I would like to say a few words and let's say about the TSVP program and really Jonas and Lynn are making a great job in organizing all TSVP participants and members. So I should be grateful to both of them for really facilitating my stay here and really the TSVP always are great environment and also thanks to Nick to setting up this great program. And I hope more colleagues will come in the next few years. So as Dave mentioned, the talk today will not be about the maritime environment, but I will make next month two specific lectures for the students, one on maritime science and another one quite related to the subject I will be introducing today. So the seminar today will be about people, the environment, human sensors and the way novel technologies nowadays provide novel opportunities for exploring understanding human mobilities is a kind of interdisciplinary research area. As Dave mentioned, I am a computer scientist involved in geographical information system now for four years, but most of the work I have been involved in over the past years and some of those works will be presented today have been developing a close relationship with geographers, with transportation planners and with a strong connection to practical and real application. What is the specific subject today? Mobilities, movement, people movements are not new but over the past centuries, of course, people have been more and more moving at different scales in space and time. And this is obviously also related to many dimensions. One of them is freedom, emancipation and many opportunities to travel in space and time. And this will offer many opportunities I will introduce them in a few minutes to understand the kind of pattern that happened in space and time, the way people move and behave at different scale at the urban scale, at the regional scale and these for many application purposes. From a modeling point of view, the idea will be to model patterns at the individual, at the, let's say, macro level to try us as computer scientists to develop modeling framework to understand how people behave, how people act in space and time at the very local level, at the macro level in order to understand these patterns. And really the idea will be, and I will browse about some of the research I've been involved in, and also before I came out over the past, let's say, 50 years and the way people especially from geographical and environmental studies have been dealing with the question, how can we model urban migration patterns? How can we model at the very local scale to the very macro from urban to regional, national scales? And I will sort of browse through the, what we call the early days, let's say in the 60s, where the transportation and geographical communities were developing massive surveys to understand how people move at the urban, at the regional scale, then a new technology and software dimension did appear in the 80s, what we call the geographical information system impact to then, and this will be the last part of my talk, how the cyberspace nowadays will offer novel opportunities, very much different from the way we were modeling there's movement, migration patterns in the early 60s and 80s. So the different dimensions I will be today discussing will be first of all, the geographical dimension where data will be made of, let's say geographical data, not only geographical data at different scales but also integration of geographical data with semantic data. And this is the way in the 80s, we were developing geographical information system to understand this mobility pattern. The second dimension which is coming up is the one of the same surveys, let's say data dimension where we do have more and more human and physical sensors that provide in real time novel opportunities to understand migration patterns still at different scale. And the last one, and all of them will be illustrated later on, is a new, quite interesting dimension coming up from social media and social network where people are acting in the cyberspace, are moving in interaction with the cyberspace and then providing a new data form that can be useful to understand this mobility patterns. Coming back to the early days, it was in the 60s and just after the second was something quite amazing, a pair for the geographic science. In fact, there is an example which is often mentioned just after the second world war, the geography department in Harvard was almost closing. And for what reason, in fact, let's say people were thinking, well, geography is dead, is a sort of static science where people are producing maps. And the question is, is it still a research area that is worth being continued at the, let's say high research level? And then came Torsten-Agestrand and this was the start of the so-called quantitative geography revolution where the idea was not only to produce maps but to consider the geographical space as a sort of a background repository space that provide a sort of modeling dimension to represent people acting in relation with different places and this still and the interdisciplinary component just appear as the center and a new way of developing socio-economical study. And in fact, the geographical domain was not very much anymore centered to producing maps but rather a way of placing the people at the center of the problem, the research problem in order to understand social and group practices. The second thing which is quite interesting especially for us computer scientists is that the modeling framework suggested by Agestrand was quite simple with a series of modeling abstraction that when, that then generated a series of F4 to integrate this different modeling concept at the data representation space. The time geography is made as you can see to the left at the top, a two-dimensional space plus time as the third dimension and where trajectories are represented in space and time, three dimension where a different constraint can be applied to for example, modeling the possible place from a given location where people can go and so on. And the next modeling abstraction that was quite useful is that Agestrand was already representing not only trajectories but also trajectory in relation with different activities people were performing in space and time. And then this gives a sort of idea of the different modeling and interaction capabilities such a framework does provide. For example, we can see at the top left for a given time we can locate the location of a given trajectory, there is a basic of course interaction. We can find for a given time the different locations of different trajectories. We can find the different path that crosses a given location. And from different trajectories we can analyze the different possible interaction in between the people. This is quite simple but this was quite successful and from this kind of modeling framework then the geographical information combination of geographers and computer scientists were then developing this kind of modeling framework where what people do, where people are acting and when people are acting. So this is let's say the descriptive dimension and then trying to make progress in relation with the different trajectories of the events that are related to this movements the processes associated to this movements and in order to make progress from the description experimentation and explanation to let's say move to work as scientists can we make some theories to understand the different reason between and in behind this evolution. So this gives a sort of basic framework and background to the different research and studies I will introduce today. This is a first example. I have been conducting in the city of Quebec where the idea was to combine the first dimension I introduced early on transportation servers and combination with this kind of modeling approach where the idea is to from the time we can set up a survey from the time we can have some data a database on what we call origin destination surveys we take a city we have oops excuse me yeah we take a city and the idea is to combine the transportation survey studies we have about a series of from a panel respondents with different ages and so on we set up some interviews GPS based data about the different trajectories plus simulation software to let's say try to simulate the different trajectories from the data they give to us and also and I will come back to that in a minute about what we call their lifetimes the idea being to analyze the trajectories at different scale from let's say the urban sale scale on a daily basis to let's say what we call the geo lifetime what people do from the time they were they evolve and so on. I will not get into the details of the way the database communities were organizing the data from best transportation survey but this is also to make a difference with what we do now with the cyberspace. In the eighties nineties when we have let's say a sort of database we organize the data with database modeling approaches we organize the data using different entities the lifelines, the time associated to the different trajectory the places where people are acting and so on and so on. So the data in the eighties nineties were very well organized and from the time the data is very well organized we have a sort of geographical information system with all the data on a daily basis recorded regarding the origin destination patterns we can have the location of the different places where people are providing the data and the surveys coming back to one of the first slides I introduced we can analyze the different patterns at the individual level. For example, here we take a family living here and the different trajectories and the daily trips from a given family using some given criteria. More interesting from the local level to the global level the idea will be to analyze some specific pattern for a given category of a population this is of course a single example is when what are doing a woman from 21 to 65 years old at 3 p.m. So you can see different patterns of activity here more interesting and this kind of query is interesting for transportation planning to organizing transportation depending on some specific family here is a single parent family employee full-time where there's people are working at 5 p.m. and where there's people are working at different hours and this provide a sort of representation of the different pattern, social pattern and we can analyze the data as you can see with different statistical techniques and then we have a sort of understanding of the urban mobility pattern at the daily, weekly scale and so on. The second level as far as it is a sort of big database we also have a survey on a long period of time about 30 years of time where we do have for a given family the different trajectories we have for example the household trajectory leaving single then in couple in single again after a specific divorce the residential trajectory how people are living in apartment in houses and so on and the career trajectory and this is quite interesting to see for example what will be the impact when there is a change of workplace of this specific kind of event moving house and we can see for example number one when there is a long woman owning a home this person is not likely to change house when we have different categories we can analyze the different patterns we can project the different pattern in space and also probably more interestingly we can make this kind of analysis what is the probability of buying a home when there is the first child which is born after three years and there is another thing which is quite interesting when we combine a different let's say population you can see that in the 60s people probably this is the let's say the period where people were let's say let's say the 60 in Canada and the US people were not very much let's say thinking about buying a home while in the 70s the green and the red and the orange people were more likely probably because of the crisis and so on having more expectation to buy the house when the first child is coming up and so on so this was before the cyberspace people were developing geographical analysis using transportation service using geographical information capability and so on and then the cyberspace came up with two dimensions I will explain the cyberspace as a whole and the sensor dimension to put things in context well this is probably known by the audience here when we came back to the 2010 most of the companies were doing things enduring things when we have a look to the 2020 most of the companies the rich one are not enduring things they are just taking data from the people and they do have a business model from the data they keep from the people meaning Amazon, Alphabet, Facebook, LinkedIn, Amazon and so on and so on and those companies from the net generating millions of data we do have other companies making a big deal about this data and us as scientists and I will come back to that in the next minutes by presenting some application we do have some opportunities to maybe hopefully to make a good use of the data so this is the thing I will try to explain in the minute the second thing is humans before artificial intelligence we are not alone we are surrounded by sensors and as you can see here we do have now more devices sensors physical sensors tracking traffic tracking pollution data tracking the people and we have the people and let's say we do have some telephone companies we do have Google tracking our data we do have much more sensors than people and the range of magnitude is of course increasing and there is an issue and there are opportunities so not quoting Stephen Hawkins but in fact he said a few interesting things before the artificial intelligence the big data is bringing a sort of revolution and us as scientists the issue is how can we transform the data as far as we have the data and provide insights by developing new mathematical tools and this is a challenge and the challenge is exciting it is about big data it is about still collecting the data finding correlation and it is also about let's say having now related to the different companies I mentioned earlier information is the new petrol for us offering novel opportunities and this brings a sort of a framework where we do have the people the environment the sensors somehow related to the title of that talk we have the old days still around where we can develop some database modeling approaches to integrate the data and then the cyberspace and all together many opportunities and the one I will be dealing with today is about human behaviors acting moving and so on but not only we do have a lot of novel areas mentioning smart cities intelligence city digital twins and so now getting into some application example a first one that has been conducted by Shoko Wakamiya who is connected by the way and she's currently now at the NARA Institute of Science and Technology it is a quite successful example of how we can infer from let's say the social data work and basically from Twitter how can we infer from the way people are acting using Twitter moving in a given space can we infer some useful information about the way people are moving in a given regional space so the idea is quite simple also research is often this is something I really believe in basic things from basic things we can derive useful data information and research Twitter in Japan the data is available is quite simple data where people are sending tweets from some locations and there's locations and there's people identify and the question is well can we have a sort of representation of a way the way a given region is pulsing and can we have some let's say inside about the way and what the people do by the way the way the people are moving at the regional space in scale in space and time the idea is not to replace the other ways of developing transportation model and so on but to have a sort of complementary view in relation to the way we can develop some spatial algorithm travel time base such as and so on but to provide something that will complement the all ways of interacting let me show the data the data is about you see the numbers are quite interesting in that region of Japan around Osaka Nagoya and Kyoto they are about 50 000 tweets per weekday 100 000 tweets in a weekend and the location of the study is mainly around Osaka Nagoya and so on so we are collecting and Shoko was collecting crowd footprints over Twitter she was extracting the different movement and from the data the idea was to generate a sort of space time representation of the different activities the people were performing a tweet is quite simple every tweet is identified so there is a data privacy issue here time is given location is given and something is said already what we can see from this kind of picture probably the next one is better at any time we have a sort of real-time pulse of the region where people are what people are sending and this gives a sort of realistic real-time distribution of the population the next one is considering of course some approximation when someone is sending a tweet and then sending a tweet from another location we do have a mobility pattern and the second interesting thing is that not only we have a sort of representation of the way the region is pulsing in terms of where the people located we can see the different main movements from the time we have these main movements we can organize space by setting up some cluster made of different caclistering algorithm we can make some using Voronoi specialization some basic entities we can derive some average time and then we have a sort of implicit representation of how people are distributed in space and this gives a sort of spatial structure of the region then the connections in between the different places the different clusters we average the time displacement and what is getting quite interesting is to have a sort of representation of the main connections and illustrating the range of things we can do is by comparing for example the different interaction in between the different places for example we can take Osaka and we can have a representation of the most connected places in relation to Osaka so Osaka is indeed a sort of hub while Imagistation is much more related to places which are quite far away so this is the sort of analysis we can have as well as having this kind of representation where we can address some time constraints for the two places for example one hour where are the places connected from one hour both from Kyoto station and from so that one is Kyoto and that one no is half an hour and one hour sorry so we can see the different connectivity in between the different places we might have a question here is well what will be the interest of this kind of study as we do have some transportation algorithm to to provide this kind of result the thing is that when there is a specific event let's say earthquake whatever typhoon etc etc as far as the data is recorded in real time we can infer this kind of pattern the other thing is well when people are sending tweets people are saying things where the second idea was experimental was well the tweet is about 200 words mostly the idea was to well can we have a sort of view of the mood of the people when they are traveling so we made some basic evaluation of the different words positive negative sentiments we made some aggregation and then the idea is still using the same database what we discover of course what people say when sending tweets is not probably what they feel but anyway we find that people were very frustrated when traveling you see very negative words this is something we can discuss with psychologists but more happy during the weekends okay some basic trend and we can see the location for example from I think the the red of the good tweets the blue of the the bad tweets we can see that for example in I think in Sakai city where people are more happy during the weekend than during the weekday and so but anyway okay the second example I want to to introduce it was from Mei and Jin she's also connected and she's now in Shenzhen so Mei and Jin did a quite related project but not this time from social media but from human sensors this is the the Microsoft Research Asia project which is called GeoLife where over a long period of time for for several years the data of a panel of population were recording in fact all the movements in the city especially when moving with taxi so the kind of things we can do still at the individual and then at the let's say global level is to make a difference for a given person the different patterns made in the morning or in the evening you can see the legend here in the map okay so we have different patterns we can have here for for example another case where we can make a difference in between the weekday routes and the weekend routes we can have for let's say two different populations the range of locations covered by this population for a given place in fact a university we can have a sort of distribution of the people connection to that university and and and so on and so on so this was a second example where this time the sensors were providing the data but still a big difference with what I introduced before and and coming back to the application Shoko did in in Japan the data was not created for that purpose you know the people who are sending tweets they are sending tweets and the data is not very much structured rate okay a big difference from the times we were setting up a big database in the 80s in the 90s and and so on the same here the data well we do have a panel that the people are happy to to have us using the data and the main difference is that we we have a lot of data unstructured data non-precise data you can say to me well when someone's sending a tweet and then sending a tweet this does not represent exactly the time of the displacement but as we have a lot of data at the end of the day the data makes sense a second application more recent that one I want to to introduce was developed by and with colleagues in Beijing and the main colleague Peng Peng is also connected here from from Beijing the idea was to this time at the regional scale analyze tourism patterns in mainland China and this time using a social data network which is in China Ma Fengyou it is a massive web social network where Chinese tourists in China are reporting their tourism activities their trajectories and so on and the idea will be well to set up a massive database from Ma Fengyou to extract the data using natural language processing algorithm and program and then to analyze the different patterns at the macro mezzo and micro scale using some network analytics the data is used as often in China we have here a representation of the data volumes generated by this application of a given period of time we do have the the market size of the different regions and already we can see with this kind of mark the sort of distribution of the different tourism activity and the idea will be to analyze the data as computer scientists once we do is to from the web social site and the reviews to integrate the data to organize the data using a network when there is a trajectory passing from different attractions we set up a network we aggregate the network and then the idea will be to analyze the patterns where are the dominant attraction are they some structural core periphery structure that appears and is there any functional motif of a local scale and the question can be well why are you doing that well the idea is to understand the patterns and maybe to reorganize if necessary in relation with tourism agencies but the different activities some details here about the the reviews which are organized the number of footprints we are creating a database but more interesting probably this give you a single example of a given footprint one user a second one we generate a simple network a second one and when we combine and of course the different connections in between the different attraction will reflect the number of time someone is passing from A to C and so on and so on and then the idea will be to to analyze the different structures is there any dominance effect we usually do that in statistics using power law functions and the idea will be is there any attraction which is dominating the other one what clearly appears that the larger the market of a given region the most dominant we the market is influenced especially in Beijing for example where some given locations do have a very strong attraction power this is an illustration of the the different patterns we do have as you can see to the top left Beijing has a very centralized structure which is also the case for Kunming and probably also for Zhang Jiang very much distributed in Bowding and and so on we also analyze at the the local level the the way the different regions and for given region the different attractions are cooperating or also some affiliation mechanism and we can see that the large of the market size for example in in Beijing there are a lot of affiliation phenomena and cooperation phenomena as well and competition is not a trend why this specific case is when an attraction is making a connection in between different clusters of attraction another pattern which is quite also a measure and obvious of course attraction is depending on the distance as you can see and also we have still for for Beijing a representation of the different attractions a phenomena in between the the different attractions on the given region and also the direction of the different attraction early on I mentioned artificial intelligence one idea we had was well fair enough we can extract some data we can infer some pattern that can be useful for for tourism studies and and so on another trend from all the data we can have from this social network is well from the time we do have some patterns from the time we can identify some social data for a given user is there any way you're suggesting some tourism activities and trajectory so this is what we we did by modeling the different dimension the spatial one the social one the different movement and the idea was to set up a sort of series of mechanism from the data which is recorded with a neural net and some probabilistic let's say derivation to find out using a knowledge graph that represents all the information we extract from the social data the tourism pattern the spatial behind information and the idea was to let's say let's say suggest some visits depending on the social data we do have about a given user the different patterns we previously recorded and so on so this is an example where a given user was suggesting this visit by thinking into account the previous data we did another study in Hong Kong still based on the same mechanism and this time it was from trip advisor it was before and during the covid times and the idea was to analyze the patterns the question being well is there any change in the way people are acting tourists in Hong Kong before and during the the covid and still by applying a similar mechanism the idea was to to extract from trip advisor this time which is available in in Hong Kong to extract the the footprints to derive some network to make connections in between the different attractions that have been part of a given tourist activity for a given period of time and and so on and then to infer some some pattern to see there is any change and what will be the pattern that emerged so still same approach deriving extracting constructing the network defining some values when there are some co-occurrences and and so on and then the patterns emerge so already we we saw before the covid the number of reviews were slowing down 2019 but of course really decreasing in 2020 with the covid we don't define one strong pattern before the covid people were 12345 visiting a lot of places but with the covid the number of places were decreasing and the the second pattern that was quite interesting is that people were visiting as you can see here we we do have a series of attraction well known which are attracting a lot of people but during the covid the places where people were going were quite different open space and much more diffuse a question that emerged from the people we have been engaged with was well what will be the next pattern other people maintaining let's say a different way of developing tourism activity other tourists just see starting to reflect there there's new trends so this is an open question this is represented here as well or you can see some strong patterns here related to the places people were as tourists traveling and we can see here a much more diffuse much more diffuse set of trends the last example to conclude I will introduce is also a recent study we we did in Austria this time not using social data but gsm telephone data where the company was providing the data so from this time billions of movement patterns given by the gsm company we were extracting the different location mobility pattern in order to analyze the change of patterns before and during the lockdown the lockdown was not tough in Austria it was from let's say a march 20 to early may what we can see is that the number of covid cases were quite huge in the west part of austria and we can see that of the trends in terms of movements aggregated in between the different regions were much more impacted in the west region and as we can see here when we make a region to region representation of the different movements the difference that slightly appear here where very much a dramatic change of migration pattern movement pattern in between the west and the east and the distances and mobility pattern that appear for all regions so the colors are probably not well represented here but in most area there were dramatic changes in the number and distances covered by the different people another one and this is also related to what I said before in the city of vienna you can see the impact of the movements during the lockdown but as you can see after let's say the release the situation is not coming back to the previous stage and this is a general and interesting question we we can have at different levels the covid probably has changed the way people are behaving the way people are moving and and so on anyway let me conclude what I want to show today is that over the past 50 years the way pluridisciplinary research is developed when dealing with transportation migration socio-economical studies in relation with some specific categories of activities have been evolved from the time of we were developing big surveys because this is this was the only way of developing structured studies and a sort of accurate representation of some transportation patterns then we we have the geographical information evolution with software coming up and a way of producing map analyzing the data and so on still with well organized and well structured approaches and then there is this big data human sensors social data coming up where the data is not generated for that but as far as we have the data we can play and and provide some interesting results I didn't get into too much detail on the way we are extracting the data providing visualization and so on but surely there are still many directions to explore in terms of combining geographical statistical analysis visual approach how to structure the data how to complement the data with confidential approaches and so on the real-time dimension is a chance because all the data can be organized and analyzed in real time is being not the case for the conventional studies another issue that can be discussed I mentioned that earlier on there's big companies keep the data google keep the data twitter keep the data so there is an issue there is also a trend which is the emergence of new systems such as mastodon which is likely to replace twitter for example we are looking for still some killer application the one I introduced earlier on from shoko wakamiya was a killer application at the time of phd we are maybe also looking for products and always a big question is about ethics people are not aware of the fact that we are using their data and this is the case for amazon which is using your data to play with the data and to infer new data this is a picture of the people that have been involved with me with all these studies most of the names are here if you are missing anyway this is about geographical human people and the environment and I hope you will have questions thank you very much great thank you so we have time for some questions anybody over there I was wondering I'm sorry could you please speak into the microphone so that folks on zoom can hear thank you for the presentation very interesting I'm very far away from what you do but one thing I noticed you on one hand used active communication like twitter and and as a representation of movement and and mood versus passive where like you have like a gps like tracking so by comparing these type of data acquisitions what do you foresee can you extract in terms of behavior you know active you know when do I decide to write a tweet is the specific human behavior versus just passively following where people go so so how how useful could this be to extract human behavior can you learn from these differences well this is a good suggestion in fact for example if we take was what Shoko Wakamiya was doing in in Japan with the tweets we might combine this kind of social data reference with additional data coming up with conventional GPS let's say data and then trying to explain the reasons what you were suggesting for someone making one specific trip at a given time also when we are analyzing the tweet messages to to understand why people are saying such things at a given time and so on and so on yes there is a range of novel opportunities when I was in Switzerland before coming here I was discussing with people from the transportation simulation community and they were saying to me well with the data you have we can probably refine the models we are developing so combining this kind of heteroclit social data but to let's say to refine the transportation models we are developing and this transportation models being based of on predictive algorithms and also some let's say well known models they do have one can help the other and probably in terms of the novel opportunities we do have combining different techniques is surely something we should explore yes right yes Nick may I ask a slightly different version of Christian's question actually so in statistical physics you have these ideas of ensembles and the things that are got it and averages and time and averages in space and kind of give you the same answers so I was I was really wondering where you have these very different ways of measuring things and very large datasets and you ultimately do a statistical analysis where you see the same patterns emerging with for example with the data that's correlated in time with tweets and the data that isn't correlated in time with tweets or whether you see qualitatively different things is it a common or a question it's a question it's it's actually a very similar to the to the last question but but maybe framed in a different way yeah I'm thinking of your question refining models we we do have by combining different dimensions here yes it could be a way for example I am trying to think about an example if we take what we have if we take what we have in in this kind of representation here if I well understand your question one direction will be from the data we extract here and from the patterns we do have from let's say transportation data and from the different models which are let's say you're usually apply one with the other can be enriched but I'm not sure I answer your question yeah maybe I can try to rephrase it so if you if you're measuring really the same thing in different ways so so when you do tweets you're effectively doing a kind of important sampling right so so you're correlating your measurement with particular events and I'm just curious when you have these large-scale averages of distributions of population flows of people etc but they really look very different when you do that kind of sampling from if you have a more neutral sensor based system let's say which which isn't waiting for you to do some social media action before checking where you are well in fact the okay in fact the results derived by this kind of analysis despite the fact that there are some approximation generally fits well the current knowledge we do have now the point is we we had this kind of question what will be the interest of developing this as far as the output is quite similar to the knowledge we do have and from the different models we are playing the answer is when there is a specific event let's say a web a massive typhoon earthquake whatsoever and so on with this kind of application you can have a sort of real-time observation of the impact of the big event on the situation and this by the way was developed by the CIA in Ukraine during the revolution because by observing the way people were sending tweets what they were saying and the way they were displacement in the city they in real time identify that something was going on so in fact despite the fact that this kind of approach is based on non-structured data with many approximation and so on the first advantage is that is free as far as we have the data it is not costly and secondly in specific case this provide a sort of real-time representation of what is going on approximation but useful yeah question up here yeah so thank you very much for the talk I just can't but sort of connecting things with with an event that is happening at OIS it's called sustainable transport arcophone I'm curious to know like if there is any part of your work that is kind of can give a contribution to this kind of initiative if you heard about this initiative because they're also discussing about analyzing data from Okinawa transportation data and yeah I'm curious to know like how your insight and your methodology could be beneficial for this kind of initiative. Well to be honest before coming here I saw the different initiatives which are going on here with the Okinawa community and I saw the GIS group of the data group here which is collecting data from the different research groups so and then I was thinking and one of the objective of the talk today is making connection of course first of all and early June I will make this kind of seminar for students but not only and the second one will be quite related but with sensors deployed in the maritime environment where we are collecting maritime trajectories analyzing the data and so on so of course I'm here to to interact so if there is a hackathon we have been doing hackathon in France for for maritime data trajectories and analyzing maritime patterns and so on so we can of course I'm still here for a few weeks so this is the objective of being here is to make connections it's not easy because the subject they mentioned that before one idea was to talk about diversity but I choose that one that does not fit completely what people do here but anyway the answer is yes thank you very much given that you have data about both people and places so you can like do a bipartite graph analysis and like see like clusters of the different groups of people or different clusters of like interesting points for those people kind of something like this yes thank you well this is a good comment too the point is somehow related to data privacy for example regarding the well there are two issues one is data privacy and the other one is the big GAFA companies are not very happy to give the data so there are two issues now related to social social data somehow is a bit more different especially from what we we did surprisingly in China with nothing war because we we do have the the social background of the Chinese people which are acting as tourists some social basic data age activity and so on and this is the way we were proposing and suggesting things but surely adding well for example in this specific case we we know nothing about the people it will be much more interesting to have additional data and combining this kind of data with let's say we probably know that I don't know in this specific region of Osaka the social trend is that one that one and then we can make some progress in not only analyzing the way people are behaving but combining with social data and so on and so on so there is an open avenue and the question is how to combine unstructured data with additional semantic data which is not given at the time and the domain is just booming you know all this twitter and facebook and facebook is almost impossible to have the data but we are in some countries and of course in some countries the data is not available at all and you can guess which country and so on and so on question yes yes thank you very much for your talk so this is related to your answer now i'm just wondering you collected data from several countries a different type of sensory data i was wondering if all of that databases are all of them have different structures or have you try to combine this one knowledge graph that somehow you know gives a different puts all the different data different types of nodes that connects together the data well this is a good point too thank you um in fact most of the data we derive and structure uh at the end of the day it generates a sort of network connection between places people acting trajectories especially and temporarily reference and and and so on we today i introduce for example the application we made in japan they are the ones in china what can we do well in austria as well cross-cultural comparison of patterns can be an idea for example uh roughly speaking uh one idea will be if we analyze let's say the way people behave before and after the covid and i will take the example of the city of vienna here there is a pattern for example and we can see that uh after the after the covid after the lockdown people were not behaving as they were before and the application i introducing hong kong the objective was different uh we do not have the data uh after the after somehow after the covid but it will be interesting to see for example how people react now uh and the way they behave and the fact that the data is structurally quite simple we are generating networks uh probably provide ways of comparing the data but next the question will be does it make sense and for what kind of studies uh cross-cultural comparison uh of post covid activities but yes we can do that in fact most of the things we we did are quite simple we take the data we extract the data we we do have a network and then we we apply some uh basic query mechanisms although we we do have papers uh of course in background nothing is really extremely complicated and this is the good good news i think people can understand what we do at least thank you all great we have time for one more question i think there was a hand up back yeah thank you for the uh for the presentation yes very interesting very thank you for the french accent very easy to understand from for many people and my first question would be about uh the metaverse and virtual environments and where people are physically in their own apartment but they can walk around in virtual places from all over the world so we're thinking do you foresee some potential good interesting questions or research paths to to address specifically in metaverse and future virtual and virtually shared environments and my second questions relate to ethics so what was your your biggest risk or challenge that you had to face during your research related to ethics and what we should all fear for the next years as the main big risks that we should face maybe in the next years regarding these questions okay three questions sorry well the first one is interesting the second one as well uh with a few colleagues we made a sort of review paper uh supporting that framework uh and the emergence of a cyberspace in geography there is the so-called first law of geography uh close closest things are more related than distant things but it's not any more valid in the cyberspace uh we can work from here with people in let's say in europe and so on and so on so nowadays in the cyberspace and if we talk about avatars replica in the cyberspace there is a new dimension and we were discussing well i'm promoting now marketing this paper we were discussing this issue in terms of how people are related through the cyberspace from the activities they perform in the cyberspace and so on so there is a new dimension here people are not related by distance but by topology in the cyberspace the way they are connected so this open uh new modeling concepts to derive now related privacy uh is a is a big and ethics is a big question uh especially as the data as human beings uh you know uh i show you some example some people are worried about that uh some people are worried about that uh and i will show you the issue okay well team bernard Lee made this statement is considered as a father the father of the of the web okay uh a few a few months ago was organized this web conference uh in portugal 70 000 people i don't know how they did it uh in portugal 70 000 participants probably some of them uh remotely they they have the data as a human being using google and you are using google as human being when i was in in switzerland uh this is the the place i came from breast uh this is ren where you are coming from and at all and google is tracking all the things i was doing during my stay in in switzerland the the places uh where i have been so google knows that uh in a specific day i was going to a pharmacy because i have something uh to deal with uh google knows that i have been making some tourism i have the same pictures from my stay here in okinawa quite similar so google knows that uh google knows he knows about you all as well as far as you are connected to wi-fi uh he knows what you're doing he was aware of the fact that i have been uh in geneva more uh crazy uh he knows because not only the data is specially referenced but also timely reference google knows that i have been to geneva and spending half an hour 45 minutes here so google knows that i have been having uh a lunch in this place so this is the way there's big data company uh are keeping the things we do the places where we have been without us knowing that you can say to google i don't want you to send the data back to me or you say to them this is what i did i want you to send back the data to me for example for example in okinawa i have all my trajectory the places where i have been with beautiful photographs and so on and so on so there is a nictix issue and google does that for you as well it's a big issue and it's a big question yeah definitely and i know that um some other companies that used to provide that information for free namely twitter have now exorbitantly increased the fees for api access for instance so you know now your average joe can't even access that information anymore i think in any country um so yeah definitely somebody to think about anyway uh thank you so much for uh your talk oh we have one more question i'm so sorry yes yeah i'm sorry thank you for your talk i actually have two questions hopefully quick um one of the questions was you had a slide at the at the beginning about you know three levels of analysis you've got the district the descriptive um uh i can't remember yeah the the question of the how and and the why yeah and i was wondering to what extent does this approach um allow you to actually access the third level explanation having inferential because all of it is you know hypothesis free to to start off with i mean that's the whole basis of big data um so to what level do you feel that you can get to a point of where you're starting to make inferences from this data that's my first question my second question it's kind of related actually is um do you feel that where do you see that this um the applications of you know this kind of approach have been really powerful um in terms of you know either social applications or something that's more translational let's say in terms of uh uh effects on on on you know human beings well in fact uh thank you for the question uh this is a good one probably in the old days uh we were very much efficient in the way we can make progress from the the descriptive to the understanding dimension especially especially what we did in Canada we were quite let's say doing well in the way we were in the position of deriving some interesting patterns because the data was very well defined extensive and so on and so on now the question is and you are right with this new big data era where we have structured data non semantically significant we are probably still stuck here you see what i mean it's difficult to move and this is probably related to to nick's suggestion we should find a way of from unstructured data and and basic trends useful but still basing to making let's say uh a path in between descriptive data to explanation and real studies and making a path toward pluridisciplinary research so there is probably a long way uh from big data from the cyberspace we can derive single things all four in the old days with very well structured costly data we were in the position of developing much more interesting things the geographers i have been working with at that time will be probably let's say not completely convinced with what i was presenting from the cyberspace you see what i mean but thanks for your question is a key one great thank you so christoph will be here through the end of june i believe so in lab five if you heard anything interesting he's eager to chat and collaborate yeah and um you'll also be teaching a short a series of short courses in the coming few weeks i believe as well yeah so sign up for those i i think i'll be attending those uh should be fun uh let's give him another hand thank you