Graphs are complicated; they come in a staggering variety of forms from your basic line graph, to scatter plots, up to hexagonal contour plots overlayed on heatmaps with tea stains indicating how excited the researcher was about the results…

Two incredibly messy and unhelpful (read: indescribable) plots
Plots taken from my office. These are extreme examples of what happens when physics PhDs get bored!

But graphs are also hugely useful tools – they can portray with a few lines what may take paragraphs to describe in full. For instance, imagine I’m holding a conker on a string that I set swinging slightly side to side. The displacement of the conker from the centre of the swing is described as a damped simple harmonic oscillator: a sinusoidal wave with a gradually decreasing amplitude.

A damped harmonic oscillator: a sin wave whose amplitude decays with time
It’s such a pretty squiggle!


To properly describe this image we ought to work out the period of the oscillation, the rate of falloff of the amplitude, the locations of each peak and trough and so on. This all adds up to quite a long few paragraphs for the describer to type and the listener to digest, which is far from ideal.

I’m excited to share that Grapheel will be working with a Computer Science student at Sussex University, Tom O’Mara, to help create an automatic pipeline to perform automatic graph description for simple(ish) plots. The end result tying in with IRIS to help give a quick initial response to an uploaded graph, which can then be expanded upon if needed.

While I’m not allowed to share the roadmap of Tom’s project, I can say that I’m looking forward to trying it out and testing it against the responses obtained when IRIS 0.2 enters beta (not long now).

Name the squiggle: IRIS and automated graph identification
Tagged on:         

Leave a Reply