Student Spotlight: Florian Krautli
Why did you decide to enter and how did you first approach the Visualizing Time challenge?
My PhD research is on time-wise visualisations of cultural data, such as timelines. Representations of events in time and ways of navigating them are some of the issues I explore. I almost had no choice but to participate in a challenge dedicated to the visualisation of time and the concept of a timeline.
At first sight, the timeline is a very basic visual layout: it’s just events organised by time. But when you start to go deeper, you realise it’s not all that simple. The challenge description outlined some of the problems: how can you show relations of events across time, how can you show dependencies and how events evolved and not just followed each other, etc.
My own focus is on representing ‘raw’ data and using the timeline format for information exploration and visual analytics. I therefore was drawn to the challenge of representing the entire ChronoZoom dataset in a way that enables a user to get an overview and discover things he wouldn’t see in the existing visualisation.
How did your idea evolve?
The basic idea came out of the way the ChronoZoom dataset is structured: as a hierarchy of timelines organised in a tree. It reflects how the original ChronoZoom visualisation looks (or maybe the other way around, depending on what came first), with timelines contained inside each other. This only became apparent to me when I examined the dataset using an online tool for exploring JSON data (jsonviewer.stack.hu by Gabor Turi). When you look at the dataset in this way, you only see the ‘root’ of the tree, the Cosmos timeline. You can then navigate the dataset like a folder structure in Windows Explorer, a familiar metaphor which I adopted in my own visualisation.
Usually my next step, after looking at how the data is organised, would be taking a first look at the data itself by, for example, producing bar charts and plots in Microsoft Excel. Unfortunately, hierarchically organised data such as the ChronoZoom dataset is a bit tricky to examine in a spreadsheet. I therefore jumped ahead and made a visualisation using d3.js and Mike Bostock’s excellent tree layout example (http://bl.ocks.org/mbostock/4339184).
This produced, quite literally, a map of the ChronoZoom dataset. You can see that the largest branch is dedicated to the history of Humanity and you can also see that there is a lot of data which is not immediately apparent in the original ChronoZoom. Some timelines caught my attention and I then tried to find them in the ChronoZoom visualisation by navigating along the map I created.
The graphical representation of the tree worked well for getting an overview of the dataset and implementing the ‘explorer-like’ navigation by making the branches collapsible was only one example away (http://mbostock.github.io/d3/talk/20111018/tree.html). Still there was a crucial element missing: there was no indication of the timespans of events and their ‘location’ in cosmological history. I therefore sketched the design below, a crossover between a tree and a conventional timeline, which is not too far from how the timelines are drawn in the original ChronoZoom visualisation. But instead of nesting the timelines within each other, I organised them like a visual folder structure. The advantage of placing them below instead of within each other is that the timelines retain a fixed height, rather than extending vertically. Therefore the writing remains legible at all zoom levels and the entire dataset can be visualised in a compact manner.
Tell us about the final project, your ChronoZoom Hierarchical Time Tree.
In my visualisation I tried to retain what I think is the crucial aspect of ChronoZoom. The tool is able to communicate the enormous timespan of cosmological history and enables seamless navigation between events that spanned millions of years to events that might have only lasted a day. It is in the design and – I would guess – the intention of ChronoZoom that one gets lost in time.
In my own visualisation I traded some of this experiential side of losing oneself in the data, for a visualisation that allowed for a better overview and also for easier comparison of events in time. The visualisation starts out with just the ‘Cosmos’ timeline (which contains all other timelines) and the first set of children timelines displayed as horizontal bands, positioned and sized along a time axis. Clicking on a timeline-bar zooms in on the time axis and the timelines that are contained within it appear. It’s also possible to zoom out again and look at the ‘big picture’ – the names of the timelines remain visible even if their timeframe would become too small to be rendered on the screen.
The ChronoZoom dataset also contains a number of exhibits. These are images, documents or videos which illustrate certain events within a timeline. In the original ChronoZoom visualisation these are displayed as bubbles and can be examined by zooming in on them. They are rich sources of information and I wanted to expose them prominently. I found it a pity that one loses all sense of context and location in time when they are examined in the original ChronoZoom tool. In my visualisation I therefore chose to separate their content from the main timeline view and display them in a separate panel.
Finally, the compact layout also allows one to answer questions concerned with comparing events. For example, would the early hominoids Genus Australopithecus and Genus Homo have been able to meet each other? The answer is yes – their timelines overlap. However, my visualisation can’t answer if their geographic living arrangements would have made such a meeting possible.
What processes, techniques, and/or tools were used or which ones were most helpful for the project?
D3.js was by far the most important tool I used. It powers nearly every aspect of my visualisation and the examples provided were of great help (https://github.com/mbostock/d3/wiki/Gallery). Especially the d3 behaviours which control the panning and zooming (http://bl.ocks.org/mbostock/2206340).
But I also want to give credit to Gabor Turi’s JSON viewer I mentioned earlier (http://jsonviewer.stack.hu/). There are a number of ways of how to examine JSON data, but I always come back to this simple tool which just allows you to paste some JSON data and examine it.
You are currently pursuing a PhD in Innovation Design Engineering from the Royal College of Art, London correct? What from your educational program ended up being most valuable in your creation of this project (i.e. a particular class, teacher, or skill)?
The project would not exist if it wasn’t for my supervisor Stephen Boyd Davis (http://www.rca.ac.uk/more/staff/dr-stephen-boyd-davis/) who got me interested in visual representations of time in the first place. He was also responsible for pointing out the potential of digital timelines as a data visualisation tool and the deficiencies of current digital timelines, especially in comparison to historical visualisations of time (one of his own research interests).
For technical help I can rely on the expertise of System Simulation (www.ssl.co.uk) a central London software company specialising in information management for the Heritage sector which, together with the EPSRC (Engineering and Physical Sciences Research Council) is also funding my research.
You write on your site that you explore "representations of time and time spans". What other explorations have you conducted, especially through your education?
I have made a few prototypes which explore alternative renderings of time that depart from the prevalent one-dimensional linear model. Another issue I explored is the modelling and representation of temporal uncertainties (see http://ewic.bcs.org/content/ConWebDoc/50996). Time in a digital context is linear, quantified and explicit as well as independent of other events in time (I wrote about Time as Data here: http://research.kraeutli.com/index.php/2014/01/challenges-for-time-as-di…). For example, all the events in the ChronoZoom dataset stand on their own and have defined beginnings and endings. In the dataset – as well as in my own visualisation – it appears as if "Earth & Solar System" began out of nowhere and then "Life" just appeared. We know that’s not the case, but how can we model and represent such temporal events?
What is the next data visualization you might be working on now or want to work on in the future?
My current focus is on the development of tools, which enable cultural datasets to be explored and analysed visually in relation to multiple aspects of time: a novel digital timeline format that respects the problematics around time as data and is able to reveal more complex relationships and temporal patterns than would be visible in a conventional format. It is directed at research in the Digital Humanities, where time is not only problematic from a technical point of view, but also poses new challenges on the demands and characteristics of a conceptual model of digital time.