August 7, 2014
by Ash Smith
The University’s Web and Internet Science research group recently had their annual “WAIS Fest”. This is a few days every year in which the researchers all down tools and do something fun and interesting. It gives them all a chance to work with people they may not always work with – important in such a large research group – but also it allows the group to try some things they may not normally be able to get done in their normal line of work. WAIS Fests are nearly always productive. In the past, WAIS Fest themes have actually become the basis of prize-winning research publications, and even started funding bids. The Open Data Service have benefited from these events in the past, which is why we always make a point of getting involved.
This year, I ran my own theme: crowdsourcing data. We’re getting pretty good at getting information from the various sub-organisations of the university, usually by showing that linking open data makes it greater than the sum of its parts. But there is some information that simply doesn’t exist in any University system, and one thing we are missing is entrance and exit information for the buildings on campus. Buildings and Estates don’t keep this information, so it’s up to us to get hold of it somehow. So our idea was to come up with a way of making it simple to curate a dataset in this way. The other nice thing about Estates not having official identifiers and descriptors for building portals is that we got to make our own, and craft the URIs as we like.
We were actually dealt a useful trump card in our quest. Our team was made up of Chris and myself, and also Alex Hovden, a WAIS summer intern. Alex uses a motorised wheelchair to move about and has had a few problems operating some of the lab doors, which are not easy for wheelchair users to operate. So when Alex volunteered to go around campus making notes of the locations of building doorways, we decided he should note down which doors he could not access, and list this in our data as well.
While Alex was off collecting data, Chris and I designed an ontology to describe entryways in linked data. It’s a lot harder than it sounds. To design our ontology we used Neologism, which is an excellent web-based tool for generating ontologies quickly, but the actual modelling process was where most of the work was. We opted for the umbrella term ‘portal’ as it nicely covers all possible entrances to things without explicitly stating entry or exit. A doorway is a portal, so indeed is a tunnel or a bridge. We also had disagreements as to whether or not security properties should be lumped together with access methods, and whether entrance/exit should be a property or a sub-class of the portal class. The task did get a bit serious once or twice as we started considering whether saying “Open Sesame” would count as an access method, and whether or not we should mention pre-1988 daleks in our description of a door with steps (we did.)
Once we had a way of representing the data, it was quite a simple task of writing a script to generate triples from Alex’s spreadsheet, and modifying the pages on data.southampton.ac.uk that describe buildings. Chris also built a smaller, hand-maintained file that lists common learning spaces such as lecture theatres, and which external door is the closest. This is now listed on the room pages, such as on this page about a lecture theatre in the Life Sciences building. Although they’re still lacking information on entry method and access restrictions, we now have a page for all the building external doors on campus that we know about, such as this page, about the south entrance to Building 2, which explains that the entrance has steps and is not easily accessible by people in wheelchairs.
Of course there are always things that don’t go to plan, and we need to learn more about certain things. One idea Chris had for getting latitude/longitude points of the doors was to leave the GPS on his phone on while taking photos of each door, and also take a photo of his feet stood next to each door with the intention of pulling the GPS information from the photo’s EXIF meta-data. This didn’t go as planned because the GPS on some phones isn’t very accurate, and although all of the geo-tagged photos were clearly on campus, many were next to the wrong building. We had plans to ask people to crowdsource spacial data by asking people to send us geo-tagged photographs, but this clearly won’t work.
The other thing we discovered is that RDF is a great way of representing this data because you can simply leave out the bits you don’t know. For example, there were several doors to which we don’t have access, and it’s not always obvious how a door opens unless you can actually open it. We were trying to list which doors require you to push or pull, which open with a push-button and which are on a sensor. It’s usually obvious which doors are on push-button, but unless you can actually see the door opening you can’t determine whether they’re manual or sensor-controlled. It doesn’t help that some things change based on time of day. For example, the doors on building 32 are sensor-controlled during the day, but become card access in the evening.
To conclude, it was another very beneficial WAIS Fest, as the output was not just a new, permanent feature for the Open Data website, but also we learned a lot about crowdsourcing, as well as the design of ontologies that describe something seemingly simple like doorways. The work we did will not just benefit lost students and members of staff, but will also hopefully benefit those with accessibility problems such as wheelchair users.