Go to ...

• HOME
• ABOUT
• SERVICES
• PROJECTS
• HELP & SUPPORT
• NEWS & EVENTS
• CONTACT

Glen Hart, Ordnance Survey Head of Research outlines areas of current and future research at the OS.

First of all happy birthday Digimap and probably the most important thing I can say this afternoon is that from an Ordnance Survey point of view Ancient Roam gets our vote. I’ll start with what the future is about really – Niels Bohr’s quote there which is well know: “Prediction is very difficult, especially of the future.” So it’s hard. Second quote by a French author: “As for the future your task is not to foresee it, but to enable it.” And basically the basis of my talk from an Ordnance Survey research perspective is our job is really to help enable that future and that’s also true as a data provider as well.

So I’m going to give 4 examples – very rapidly in 5 to 7 minutes – each of which I think shows a different example really in terms of data supplied for the future. So I’ll start with something called the link data web which you may or may not have heard of. Probably about 6 or 7 years ago in Ordnance Survey we started a research theme in something called the semantic web – and this is really one of the more active areas of semantic web generally nowadays, its something called the linked data web and it’s based around the idea of people asked a question about 3 years ago about data which is why can’t we represent data in the web in the same way that we represent web pages. In other words, we interlink data. But when we talk about interlinked data we don’t just mean at the dataset level, we mean at the individual item of information. This is becoming quite a ‘thing’ at the moment, certainly in government initiatives now to try and push data and the publication of data in the link form.

So what do I mean by link data? Here’s a very simple slide to try and get across the ideas – so there are 3 data sets here. Maybe there’s one here which is a Real Ale guide or something – something quite dear to my heart, maybe a business directory like Yellow Pages, and then some geographic information from say organizations like Ordnance Survey. And essentially we can just link bits of information together over the web – we publish this and then we put links in to that data. I won’t dwell on that – but it’s not quite as simple as that, but it’s almost as simple as that.

So what we’re doing at Ordnance Survey – we’ve already published an administrative gazetteer in linked data format – there’s the address if you just type in data.ordnancesurvey.co.uk you’ll get access to that. What that is essentially is a query-able data set, that you can just navigate by clicking it as well. So it doesn’t look any different, but it’s what’s going on behind the scenes which is quite different. We’re developing a place gazetteer as well, and we’ll go from there. One thing that is worth saying about this is it’s a different way or presenting the data – it’s a different way of interacting with the data – but it also creates new data as well because those links themselves are data, and they are as valuable as some of the data that they’re linking.

So what difference will linked data make? Well it will change the way we think about data, it makes the physical task of integrating data must easier- but not the intellectual task. So, we still need to think about what data we’re linking together and why we’re making those links. Because it creates explicit links, then those links have value. I’d like to say that it’s probably one of the most fundamental changes really in the way that we think about data and present data, and we really don’t know where it’s going to lead. It’s at the start of something very, very, different really. It could be exciting.

I’m going to touch a bit on 3D and height. We’ve done quite a lot of work over the last few years, but what we’re trying to achieve now is a picture where we’re building an integrated solution – the research behind this has been largely looking at how we can economically capture more detailed and more integrated height information. And I’ll just run a series of slides, and as I’m running these slides, this is just showing, if you like, the way we build up our current 3D data solution. This is an area of Bournemouth -Bournemouth is very popular with us for 3D. We’re starting off with photogrammetry as much as anything else, classifying the image – so this is working out where vegetation is for example - and then fixing it against our OS MasterMap. So we’re actually integrating it with OS MasterMap and that’s actually important to us, and ultimately building a terrain model as well. So what we’ve essential done is built a model which integrates the terrain, MasterMap, and 3D buildings. The sort of questions that we’re asking are – well, how complex can we economically make those 3D buildings? There’s some quite complex ones there and some very simple single building height model ones as well. But even with single building heights one of the questions is really ‘what is that height representing?’ Is it the very top of the highest point of the building? Is it the eve height? Is it the surface height where it hits the terrain for example. We’re using this data to try and investigate what customers potentially want out of that. What we do want to make sure is that it’s integrated in the sense that we’ve got those 3 components integrated together and fit together, it’s maintained and maintainable, and also linked. Because potentially, this could be exposed in a linked data web or it could be exposed in terms of linking to the other components that Ordnance Survey already provide.

Another area which could have a big significance in the future is automatic map generalisation. Again this is an area we’ve been working on for a number of years, but has started to show some sort of pay back more recently. So we’d start with OS MasterMap and ultimately turn that into 25k mapping in this case. Very much the case at the moment that that’s very manual. It’s technically difficult to do that automatically – well it’s technically very easy to do it automatically it just looks rubbish if you do it simply. So the trick is ‘how can I produce something which is usable at the end of the day, completely automatically’. And that’s what we’ve been working on. These are some of the results that we’ve got today – untouched by human hand. A lot of people say well, on the survey are possibly not human? Certainly in the research department it’s debatable. So these are just different – because it’s automatic – these next slides will demonstrate ‘well what can I do with that?’ because it’s also a vector. Well it means we can put different things on those maps, take things off, colour it differently, we can say selected land use on but no use – that should say land cover, sorry - or little land use names and some land cover, put some contours on there. So really it’s about producing mapping that’s a lot more flexible for the future. It will be vector based and it will enable different map variants for Ordnance Survey to be produced relatively simply and quickly, which you can’t do it you’re relying on manual systems, vector enables user customization, and it’s also more economic to produce. The end benefits to all of us are really therefore are you’ll end up with something that’s more usage, more flexible, and it’s more affordable.

And lastly, I wanted to touch on something which is not really about end products in the physical form, but it’s about what we’re trying to do in the way we approach building those products for the future. And we’re adopting something called user centred design – you might ask yourself ‘what is user centred design?’. In one sense it’s fairly self evident but as much as anything else it’s to prevent things like this…and things like this. The first version is something which clearly doesn’t do what the user wants it to do. And from the manufacturers point of view it’s also quite bad because there’s very few repeat sales. It’s probably a good thing if you want to stop wars – it actually has most of the components that the user actually asked for, it’s just they got one thing basically wrong. In this case it actually does do what the user wants but it’s not very easy to use. And the challenge is to apply that to data. Now traditionally there has been a lot of work done on this, in particular in areas like software but also engineering generally. A lot of that work in the software field would be on user interfaces. This isn’t about the user interface, this is about fundamentality how we go about designing the data. So there’s almost nothing done on that field so far – we’re collaborating with a couple of universities that have done something in the same area, but there really aren’t many people working in the area at the moment. And it’s not about user interface – because there isn’t one with data directly. It’s about what the data contains, how the data is structured. It’s about how we package it and how we supply it, and a whole number of other things as well. So it’s really about taking these ideas and coming up with ways of actually working out basically what’s bad about the way we do our data at the moment, and then how do we fix that? And how do we do that in a systematic way – so we’ve got design principles that we can apply. So in result we’ve produced more usable products, lower cost of use, lower cost of support, basically which means overall – lower cost.

For a sort of conclusion – I think for a data provider anticipating future needs, that hard bit about predicting the future because we still need to do that. But there’s also something about developing that future ourselves as well – actually doing things actively to develop a future. It’s about developing new content, new ways of organizing the data and new ways of interacting with that data. And those are some of the things that I hope I’ve shown some of the activities that are going on within the research department in Ordnance Survey. And last a shameless plug for if you want to find out more about linked data – I’ll leave that up there for a few seconds – this is something we’re running in March.

I’d like to finish on one final thing - the original question was ‘what’s the future of geo?’ And in my opinion the future of geo is the future of information in the sense that geo has to be part of the whole information field - it can’t be something seen as something a little bit outside which traditionally it has been.