Why Are Topic Maps not Maps?
While the choice for the term Topic Maps is quite natural to me, it was (and still is) not yet clear to me how far the analogy with maps actually goes. There is something about cartographic maps which all graphical representations of topic maps I so far do not seem to have:
- A metric, i.e. a concept of distance.
- An elevation, or more generally, a spatially varying property for the 3rd dimension.
I personally miss these concepts from Topic Maps and specifically from graphical representations of maps. Conventional visualizations are all for graphs, and the only issue there seems to be which details of maps to show and which not:
- in which detail should associations be shown?
- should typing information be included?
- should only a focus of the map be visible?
This is all well and good but there must be a reason that this eye candy often remains exactly this: eye candy. And as such these representations are difficult to navigate with. Too many dimensions.
Let us have another look at conventional maps.
Any distance between two points on the map corresponds to the geographical distance the real locations have, depending to a more or less high degree on the projection function which is used to connect every location on the 3D sphere (Earth, for instance) with a point onto a 2D plane.
One obvious choice of distance for Topic Maps is to use the distance indicated by the number of associations between two topics, but that gives us the same topology as graphs.
There are many more problems with this, one being that all associations are counted to have equal distance. Which does not really reflect the semantic distance between two topics.
Another more linguistic pathway would be to find the semantic distance between the concepts the topics represent. But that is at best a research field and would imply that there is a lookup database for coordinates. And these coordinates must be absolute, so that it is possible to relate every point to every other point. This does not seem to exist yet and if it would, it would somehow question the claim of the TM community to bring mapping order into the concept chaos.
There are actually a number of clever ideas how to make graph-like representations more meaningful, among them being cartograms and TreeMaps. The following map shows the death toll of avalanches between 1975 and 2000:
(The Westbank in Palestine seems like a safe place to me.)
Accordingly, the space a topic occupies is somehow proportional to ... well, any number we generate for that topic. That can be the number of associations it is involved in, the number of occurrences (or even names), or any combination thereof.
It can also, and that is much more meaningful, be derived from the documents the occurrences point to. Sure, it would mean that these documents have to be analyzed for the relevance of their content, essenced and statistically processed, but that is nowadays quite trivial.
A graphical map representation would then look more like the cartogram above. Many researchers have been here before, such as for instance BibTechMon.
But what about a third dimension? Something like this:
What should it actually mean in terms of the topic map content? What are high mountains, steep slopes and long ranges?
More about that later.