Cartographers can’t be trusted. It’s a controversial view (possibly), but I feel like they’ve always kept to the furtive fringes of society – happily knowing exactly where the fringes are – and they’ve done this for a reason.
First off, maps are difficult to fold.
Secondly and worse than that, they’re also full of lies. I’ll reluctantly admit that they’re the good sort of lies. They’re the sort of lies that enable society to function and get people home. But is that enough?
Of course there are lies that we all tell ourselves to get things done. But I’m currently trying to describe the world (via an ontology), and the realisation that places don’t really exist came as a bit of a shock. If this is the first time you’re hearing about this cover-up then you’re probably furious too – or possibly dubious. So try this:
Imagine what would happen if Whitby and Buenos Aires swapped names.
Unless you’re enthusiastic about nominative determinism then you’ll be hard pressed to think of any change that will occur to the things that the names describe.
Names are most useful when they describe things that are fairly reliable. People, even bigamists, are this sort of reliable. Even though we age and grow and change a bit, we’re still the same us. Choose an even trickier example (a thing that’s even a type of place) – rivers. Heraclitus told us that you can never step into the same river twice. But most people are OK with the idea that a river has a single name. It’s not the water that makes the river – it’s the decision cartographers took when they gave it a name, and when we all agreed to believe them.
Names, in fact all words, are only meaningful if there are enough people willing to share the definition. On the semantic web, this type of cultural contextual complicity is really difficult. Computers only know exactly what you tell them, or what can be directly inferred. The problem with places is that we use the same word to stand for a location at lots of different times. It’s the exact opposite of foreign languages. Translating is relatively easy. We have a single thing/idea and various alternative words that all mean roughly the same thing. With places, we have lots of different places (temporally speaking) but we tend to use the same name for each instance – we don’t define the context.
Take Rome for example. Think of Rome and you might ruminate on Romans ancient and modern. Maybe you’re religious and you pondered the pope for a while. For people, the lie that places exist works perfectly. We know that Rome has changed. Buildings, people and probably almost everything else has come in and out of existence – been true, false or just a Roman rumour. This is fine for human being with our big, beautiful brains. We can usually detect the context and which Rome we’re referring to. But computers can’t, not automatically. So we need to provide the context. We need to work out why the mapmakers lies work, and how we can manage to make a computer do the necessary fact-fiddling to work out the context.
There are loads of different types of lies, but the sneakiest type are the ones where you’re lying by leaving out facts, rather than providing false ones. These are the lies that cartographers commit (and politicians).
Maps describe places, people think. But they don’t really. They describe spatio-temporal entities. They’re representations of a place at a specific time.
As a cynic I used to think that mapmakers brought out a new copy of their map every year just to make money – now I think it’s probably a pre-emptive move against the legal repercussion of their lying. Maps are a snapshot – they might make a nod to this in their annual editions, but they’re still pretending that places exist, which is why I hate them, all of them.
So this brings us to the proper purpose of this post. I think I might be a mapmaker.
Ontologies are maps of the world. They’re maps that computers read to make sense of things on the web and in the world. The future of the web rests in our ability to make statements about concepts and resources, and the ability of computers to understand them. They can then create HTML and RDF aggregations based on embedded Linked Data identifiers, ontologies and associated inferences. Ontologies are the A-Z of the future web. I’m creating an ontology to describe places.
To be fair I set off with the best of intentions. But however much I tried, I soon found myself lying like a mapmaker. I now think there are two ways to describe a place:
1. The mapmakers gambit – basically this is pretending that places exist. It works, but it limits what you can say about these things. The map is only true until something changes and we’re not talking about changes to the physical reality, we’re also talking about a place existing in multiple contexts at once.
Concepts are in constant flux, so a single description will never be accurate. Just think of Glastonbury – is it a festival site, a farm or a druid hotspot? Most domain-driven design is concerned with fairly limited domains – they know the context. The problem with places is that it’s a very broad. The mapmakers gambit sort of works, but it isn’t very accurate and it limits your ability to describe the world.
2. The time traveler tricksibelle – there’s nothing quite as satisfying as doing something right. Places work in everyday conversation because the context (temporal or otherwise) is implicit – we travel seamlessly into the correct context. In an ontology you need to make context explicit.
You can do this by creating contextual entities that you can relate to statements about places. One example is a spatio-temporal entity. These little pockets of time can share a location – and are probably the thing that most people think of as the place – they’re the combination of place and time. In other words you create a place called River Thames, and then you can tell the story of the place by creating spatio-temporal entities for each significant change in the context.
Of course, we’re then faced with the question of what constitutes a context change, and what is a place change? How do you decide when something has changed enough or it to constitute the creation of a brand new place – when is the context created by the spatio-temporal entity not enough?
A good example is Wembley Stadium. Wikipedia has two things for Wembley Stadium:
- Wembley Stadium, the current stadium, opened 2007 on the site of the 1923 stadium
- Wembley Stadium (1923), the former stadium, built in 1923, demolished in 2003
That’s something I’m still working on – maybe I need the advice of a good cartographer.