Using sound to represent uncertainty* in address locations *positional accuracy Created an extension to ArcGIS Evaluated this using 49 users of spatial data - Comparing Visual and Sonic methods
Before we look at why I’m using this, I’ll explain a bit about Address Layer 2. What is AL2? Point for each postal address (e.g. 11B Clarendon Road, Norwich, NR2 2PN) Contains lots of information, as well as the postal address and spatial location Links to MasterMap Topo using TOID ( Topographic Identifier ) It also contains some uncertainty information / status flags Data from Norwich MasterMap & AL2.
Uncertainty can be very important Depending on what you are doing with the data AL2 is often used for geolocating data, and subsequent analysis Also for routing – Cornwall example. AL2 contains 5 different status flags Including Positional Accuracy. Which represents the accuracy of the spatial location of the Address Point with relation to the building the address refers to. Which some users ignore – according to Ordnance Survey Staff. Why? it's not relevant users don't think it's relevant (when it is) users can't access the information within the data users can't display the information the information isn't available Am focusing on one issue here – but need to remember about the others. Data – AL2 Technical Spec & OS Usability Document Census Boundaries – AL Coursework 2, Health & deprivation practical. CACI Acorn -http://www.igd.com/index.asp?id=1&fid=5&sid=45&tid=73&cid=481 & CACI Acorn Brochure - http://www.caci.co.uk/acorn/downloads/New%20ACORN%20brochure.pdf Google Maps – 2010
Vision can be saturated Finnish Town & London Crime (Harrow, Stratford, Westminister) So not easy to show uncertainty Alternatives to Vision There are ways of using vision more effectively – but there are times when this isn’t possible Sound is next most powerful sense Little Previous Research Fisher (1994) Uncertainty and Soil Type MacVeigh & Jacobson (2007) Harbour, Sea and Land There is a need for the ability to represent sound in a industry standard GIS, potentially utilising the ability to create custom scripts (ArcObjects) as extensions. These were pilot studies, custom coded. Need to move these to a generic, easily usable environment. No user testing completed No existing research frameworks Piano Notes CEG Triad, was the preferred option in the pilot study Easy to understand – good order. Possibly could have chosen anything though. Very little existing research
AL2 Positional Accuracy Categories Positional Accuracy (PQA) is what we’re interested in. It represents the accuracy of the spatial location of the Address Point with relation to the building the address refers to. Create an extension to ArcGIS Why ArcGIS – easy & what I know Specific requirement for sound to be added to industry standard GIS Prototype Pilot studies at Ordnance Survey & UEA Lead to improvement of extension and questionnaire Reduce 5 categories (sounds) to 3
49 participants, 19 OS, 23 UEA, 7 LA All used GIS and spatial data on a regular basis Two Stages Computer Based Evaluation Presentation Methods Demo Visual Sonic Visual & Sonic Same Visual & Sonic Different Discussion What participants thought about the software and sound More Qualitative results Qual & Quant
Two Main Factors Proportion of Data 25% and 75% easier than 50% As you might expect with geographical data in a visual way. But couldn't find any GIS literature on this. Small amount of research in medical field, looking at how people estimate proportions of stained tissue samples. Presentation Method Visual Sonic Same – best, better than visual only. Visual Only Sonic Only – fairly bad, but similar to VS Different. Visual Sonic Different Logistic Regression To see how the two factors interacted (as it was a combination of the two, rather than one or the other) Also Knowledge of the Data appeared to have an impact – but when this was added into the logistic regression, it wasn’t significant. Discussion Sessions Greater difference between sounds Either different instruments, or completely different sounds. Colour blind users Could be very useful for them, but too small a sample to test effectively.
Summary Sound can work Proportion of Data and Presentation Method - were significant Knowledge of Data - is important so something I need to consider Improvements Task may have limited the wider applicability of the results Possibly use more complex task (e.g. clustering) and/or different sounds Next data UKCP09 – probabilistic climate change data VR Visualisations of Future Landscapes – uncertainty of particular elements in the visualisation
Using Sound to Represent Uncertainty in Address Locations Nick Bearman and Andrew Lovett PhD School of Environmental Science University of East Anglia [email_address] ; [email_address]
Methods Value Description Surveyed Within the building that the address refers to. Approximate Usually within 50m. Postcode Unit Mean Mean position calculated from correctly located points within the postcode unit (e.g. NR4 6AA is a postcode unit). Estimate Usually within 100m. Postcode Sector Mean Mean position calculated from correctly located points within the postcode sector (e.g. NR4 6__ is a postcode sector). Value Surveyed Approximate Postcode Unit Mean Estimate Postcode Sector Mean Introduction Why AL2 & Uncertainty? Why Sound? Methods Results Next Steps
Participants asked to choose proportion of values that had AL2 PQA Value of Surveyed (25%, 50%, 75%)
Introduction Why AL2 & Uncertainty? Why Sound? Methods Results Next Steps None AL2 Positional Accuracy Visual only AL2 Positional Accuracy Council Tax bands Visual and Sonic representing different variables AL2 Positional Accuracy AL2 Positional Accuracy Visual and Sonic representing the same variable AL2 Positional Accuracy Topography outlines only Sonic only Sonic Data Visual Data Presentation Method
Also Knowledge of the Data appeared to have an impact – but when this was added into the logistic regression, it wasn’t significant.
Greater differences between sounds
Colour blind users
Results Factors added to Model -2 Log Likelihood Cox & Snell R 2 Proportion 182.01 0.043 Presentation Method 169.579 0.105 Knowledge of the Data 167.319 0.116 Introduction Why AL2 & Uncertainty? Why Sound? Methods Results Next Steps 0.88 75% 0.62 0.74 Mean Score Proportion 25% 50% 0.82 Visual 0.65 0.67 0.86 Mean Score Presentation Method Visual Sonic Same Sonic Visual Sonic Different