 We'll do the three presentations. We'll make sure to take a break in between each one to get another drink. The way we will decide the winner is by your applause. You're ready to take your way? It's all yours. This is the data source that's being visualised. The Australian Women's Register. It's a data graph of biographical records about multiple women, so those are people who have had archives about them. Other local state themselves were artists or politicians or writers. Or they're famous enough to have had people writing about them that have fed what they represent in the archives. This data set is published by the East Goldship Research Centre at University of Melbourne. And it consists of a number of about 5,000 XML records. It's published also as a website, but this visualisation is based on XML records that have been harvested using the Open Archives Initiative protocol from the data harvested. Firstly, we use the Open Archives Initiative protocol to extract the 5,000 XML records from the server at the ESRC in Melbourne. And then I wrote an XSLT that applied to each one of those records, which selected out just certain pieces of information from the records, which are quite rich in terms of biographical data about those people, and I was looking at only a few of the important relationships so that I could visualise those things. So the XSLT converts those XML records into a year of XML, about 5,000 XML data graphs. Those graphs are then pushed into a sparkle store, a triple store, to produce an aggregate of all of the records at once. Then I write a sparkle query to extract the data out into a table of form, I load that table, CSV file into the SytoScape tool, press the visualisation button, and it produced a fabulous visualisation, which was fantastic, which was great because most of the work the agent isn't preparing the data, and as it turned out, the actual visualisation tool itself was a dobble, it was a snap. So, about the data set, as quickly we've got people and groups, so the individual people are, you know, notable women. The groups are often like feminist groups, they may be political parties, that sort of thing. The individuals also have named occupations, which are either they might be paying jobs or they may be, you know, applications if you like. In the visualisation, the people, the groups, and the occupations are all represented as nodes, as the dots on the graph. And the group membership is indicated by purple lines, that link a group to its members. The occupation membership are green lines between individuals and their occupations. So the interesting thing about the visualisation then is the way that those relationships end up producing a graph which related occupations ended up being situated approximately on the graph. And occupations and groups sort of move around to produce a graph in which overall the tension of the graph is minimised. And if you like, all of the little nodes are kind of, they are repelling each other as if they are statically charged. And what's holding them together are the relationships between those nodes, which are like stretchy rubber bands. And there's two competing forces which end up producing a graph that you'll see. So interesting things to note about this graph, if you zoom in, you can have a poke around and see what these nodes actually are. This cluster of occupations here, these are all the people who have a common occupation, which is the node in the centre. Those are political, those are parliamentarians. And so it's interesting to note how distant the politicians are from the rest of the population. An interesting other group of allies around the bottom are sports people. These nodes represent gold medal winners from Olympics and Commonwealth Games and things like that. Also particular types of sports. So there are swimmers, swimmers and cyclists and netball players and things like that. Again, they're not very closely tied to the other notable women. In the graph you have a much more dense and interrelated set of nodes. These are journalists of various kinds, TV journalists, radio journalists and so on. Around the other part we have social working people, medical nurses, social workers and so on. So very interesting to see although I haven't spent a lot of time yet exploring the graph, very interesting to see that those social relationships end up being represented especially. Fantastic. Well done. Good afternoon everyone. Thanks for coming. I'm Adam, this is Josh. We're both from Perth. Don't hold it against us. We're programmers at CSIRO and I guess the shortest way to describe what we do is to try to help link scientists and data. So yesterday we were introduced to Janet Newman who's a crystallographer also at CSIRO. Crystallographers for those that don't know basically try to create crystals out from combining different combinations of proteins and chemicals in tiny little droplets on trays. The formations of these crystals is unpredictable and it's difficult to know what combinations you need to create crystals so the crystallization process isn't always consistent. Some conditions are more conducive to crystallization than others. There's been a lot of work done these are the examples of the droplets here. There's been a lot of work done in identifying the crystals like this out of the data sets. These are the ones that haven't crystallized. What we're interested in looking at is trying to see if we can find relationships between these un-crystallized forms to better understand the process and thereby help create more reliable methods of creating crystals in the future. We were given a few thousand of these images along with some pre-rendered masks. This is the mask here basically this just identifies the area of interest on the subject picture. The black stuff we're not interested in when it's overlaid there that helps us eliminate any noise so that we can just focus on what we want. We looked at several algorithms to do this. The difficulty we had is that it's very hard to identify specific features of interest so there are certain characteristics like the darkness or lightness different geometric properties similarity could be any number of those different things or any permutations or combinations of them. These are examples here of the software that we've got so the idea of this software is that Josh has selected a target picture and we're basically saying we want to find other pictures that look similar and as you can see it's done a pretty good job it's picked pictures that show this crystallization occurring there are other sample sets this is really the real use case so we've picked an image that doesn't have crystallization and we want to say show me all the others that look like this and it's done a reasonably good job of picking out similar things I'm not sure if that might be showing crystallization there I'm not 100% sure but there are issues with it it's not perfect. The last example shows where it really starts to struggle in this case we've got an image that doesn't have crystallization and it's got a bit confused and picked a lot with crystals but we think it's a pretty good proof of concept we've found some white papers discussing more sophisticated similarity techniques but it's difficult to implement those in two days so we think it gives us a push in the right direction though we've had good feedback from Janet who says that this kind of thing is going to be useful I guess we're thinking about trying to do something a little bit more sophisticated with the open CV library that's open computer vision library but yeah hopefully it's going to help someone create more crystals. Thank you very much. Fantastic. Please welcome out in the third and final competitive entry Thank you. So this is a butterfly project which is exploring the fundamental interconnectedness of all things so I was interested in the butterfly effect which from chaos theory as described by Ed Laurence is where small changes in one part of a deterministic nonlinear system can have large changes elsewhere in that system so for example a butterfly flaps its wings here and halfway across the world causes a hurricane or as Douglas Adams puts it what we are concerned with here is the fundamental interconnectedness of all things the connection between causes and effects are often much more subtle and complex than we with our rough and ready understanding of the physical world might naturally suppose so in order to explore this idea conveniently we have access to a global network of sensor data streams essentially observing the very small changes that occur locally and I thought how can we make use of those data streams to identify patterns or correlations between seemingly unrelated streams of data so what I've done there are three parts to this project the first is a sensor unit which is here so this is an electronics project and what I've done is taken a DHT-11 temperature and humidity sensor which is fairly standard sensor unit and connected it up to an Arduino microcontroller and a program that's microcontroller using Arduino wiring language to hold a sensor every minute so this is the metaphorical part if you like and what it does is it publishes the data via a serial connection to the second piece of the project which is a Java application which listens to that serial connection, parses those CSV values and then pushes them up to the cloud to a service called COSM and COSM is basically designed for publishing data streams from the internet of things so then we can take that stream of data and other streams of data from unrelated sensors or similar sensors all around the world and aggregate them into a data visualization so if we take a look this is the sensor data stream that's being produced by this butterfly running on COSM and I might actually just pop to a separate window here so this is you can see we've got graphs being generated over time it's remained pretty constant recently but you can see earlier on I was blowing on the sensor to make a change and there's also a changing and the due point has changed just recently but there are other people who publish data streams for example this is a stream that I discovered through the API of COSM in Canberra which I don't even know who's running that but I can make use of it and we can put those together into something like this data visualization which is a very annoying visualization that we can then use to explore the data and I must say that this is actually running on just random data I didn't have time to integrate any meaningful data sets into this but you can see the idea of what we would do so it's my hope that as the incident of things grows and as the amount of sensor data available online increases that tools like this can contribute to an improved understanding of the nature of the fundamental interconnectedness of all things right so I'm now going to ask my judges to come up and each judge is going to defend very passionately why they think you should actually vote for the contestant they're representing Connell why should you vote for Connell Connell in a three minute presentation to the judges produced knowledge from data you know and you can't argue with that he had a beautiful visualization which actually showed knowledge from data and that's what he researches about the interesting contrast between what these guys have done versus what all the other contestants were doing is all the other contestants went back to their own knowledge and built from that these guys were clever enough to scar the rest of the world for what other clever people had done and built on top of that and I am a total proponent of reuse and I'm afraid that's the only way that a relatively small country can beat its way up against all of you yeah if you reinvent everything you end up with the wheel in five years time it's already there that's why these guys were absolutely one so I think Ana really really touched on the point that she didn't even really know where the data was coming from and that really exemplifies what this project is all about it really touches on creativity standards, reuse ease of use it hides the detail of what you're doing with the front end away from the user the researcher and it's an idea that can scale to be absolutely huge to look at potentially millions of data feeds from around the world it took her 24 hours to build that okay and that is really really good because it builds on existing frameworks existing standards the Arduino etc etc it uses the nectar cloud which is a big tick for us and it automatically plugs into real-time feeds as I said potentially millions of them and it can scale out and really this whole idea of exemplifying you know the networks the neurons it's a it's a sensing of networks, connections into our minds to examine how we can look at the world in a much better and more effective way so I think that's why this will be for Ana in 3, 2, 1 I think everybody knows the winner Ana come on Ana congratulations congratulations can I be a fun event I have for that