What does rising global carbon dioxide sound like? Or the crash of the pound? How about Sydney’s record-breaking rainfall, or the share value wiped out following Facebook’s pivot to virtual reality?
While all of these things have been frequently graphed, now we can turn them into audio as well.
Introducing noisycharts
Noisycharts is a new tool created by Guardian Australia to easily turn data into sound, with an animation to accompany it.
Here’s five examples which demonstrate the features of the format, with more details on the project and what it will be used for below.
The first noisychart shows the decline in Mark Zuckerberg’s average net worth per quarter over the last four quarters. This chart demonstrates the sampler mode which can be used to map data to actual instruments and other noises – in this case, I’ve built in a “Sad Trombone” mode.
Meta, the parent company of Facebook, has had significant drops in its share price following the company’s pivot to virtual reality, as it also faces pressure from competitors and poor market conditions. The lowest tone is US$36bn, and the highest is US$63bn:
Click here for a version of the chart with voiceover
In 2022, Sydney, Australia has had the most rainfall of any year on record. This noisychart maps the cumulative total of rainfall over the year to the pitch of the tone to show how the record was broken. The lowest tone on the chart is zero, the highest is 2,364mm, and each note is a day from 1 January to 31 October 2022:
Click here for a version of the chart with voiceover
The sampler mode is well-suited to novelty charts, or matching sounds to the content. In this noisychart, you can hear how the cavoodle has risen to prominence over other dog breeds, with the barking pitch mapped to the dog breed’s rank in popularity in Australia. The lowest tone is a rank of 20, and the highest is for number one, with each note a year from 2008 to 2021:
Click here for a version of the chart with voiceover
Despite how well-suited the tool is to creating novelty charts, it has a place in covering serious topics, and making them more accessible. Here, I’ve mapped the value of the British pound in US dollars to pitch, so people can hear the dramatic drop as the pound fell to the lowest value since 1985 following the Kwasi Kwarteng and Liz Truss mini-budget. This chart also demonstrates the “fully accessible” mode, which is explained further below:
And the final example shows the most urgent issue of our time – the rise in carbon dioxide that is driving global heating, along with other greenhouse gases. This noisychart demonstrates a different tone to the others, which makes it much less mellifluous. The lowest value on the chart is 28m tonnes, and the highest value is 37bn tonnes. Each note is a year from 1800 to 2021.
Click here for a version of the chart with voiceover
Why did I make this and what is it for?
In short, I want to make charts we use at Guardian Australia more accessible, and bring our data journalism to new audiences on other platforms.
Importantly, one aim is to make our data journalism more accessible for vision-impaired people.
During the pandemic we tested an early approach to data sonification in a few of our coronavirus charts to make them play audio – that is, to map the data against an audio frequency. This was a prototype to explore allowing people using screenreaders to play a chart, rather than read a description from the alt text:
Here is an (experimental) audio chart of NSW case trend. Some ppl might remember the prototype for this a while back? Now we have built it in to our chart tool, and have switched it on with some line charts for testing #datasonification #dataviz #d3 https://t.co/bmIS9aQT6r pic.twitter.com/zT0qkctSbb
— Nick Evershed (@NickEvershed) September 1, 2021
While the early version was functional, it didn’t sound very nice and actually wasn’t that accessible (the play button was basically impossible to find with a screenreader, for example). However, the concept was promising, and the very small animation we did with it initially suggested that the audio chart might also be of interest to a wider audience.
It was clear that we needed to refine the concept, and also test different approaches to data sonification to find the best method for representing different types of charts and data in audio.
The other reason I made the tool is to give our data journalism a wider audience. Representing data in podcasts using audio is rare (but it does happen), with presenters often preferring to describe a chart rather than create the necessary audio with existing data sonification tools. In video, charts are often either flat graphics with no animation and sound, or require hours (and sometimes days) of work using video production software to animate. And while there are some great options for making quick animated charts, these have no audio.
The rise of video-based social media platforms is also a consideration. TikTok focuses exclusively on video, and Instagram has recently chosen to prioritise video over other content. As such data visualisation is not common there.
It’s my hope that lowering the barriers to producing customisable, shareable, animated charts with audio will help lead to our work being seen more widely, and bringing greater clarity to our audience.
This is not the first tool that has been built for data sonification, and is not even the first web-based tool focusing on data journalism. However, noisycharts has a few differences in approach to other projects:
It very quickly generates output that is shareable in video, on social media or in an audio recording.
It has a mode for accessibility built in, which uses text-to-speech to introduce the scales and data series. It has other modes in which the scales would be written in the text of an article or social media post, or introduced by a presenter.
On the technical side it is built to work with D3.js, the industry standard Javascript library for datavisualisation. This is done intentionally as several of the most popular chart-making tools use D3, so I’m hoping this work might lead to the accessibility side of the project being replicated elsewhere.
Overall, I’m hoping noisycharts will help us figure out the best approach for the audio and text to speech features in the accessibility side of the project, which we’ll then incorporate into our in-house chart tool. It also means we will have a new tool to produce dynamic, animated charts for video, social media and audio reporting.
It is still very early days for noisycharts, and I’d appreciate any feedback on the project – particularly with regards to the roadmap below. You can contact me at nick.evershed@theguardian.com.
Things I’d like to see happen with the project
Partner with accessibility researchers and advocates to figure out the best default settings for various chart types and incorporate this into Guardian Australia’s charts.
Explore open-sourcing the data sonification module and encourage web-based chart providers to explore using sound for accessibility.
Expand the features and formats, eg include scatterplots, vertical bar charts with categories, and polyphonic audio playback.
Experiment with mapping data to other variables, such as tempo, stereo position, and other synth parameters.
Develop more options for animation – line chart races, animated labels, optional transitions, colour schemes.
Pilot a podcast or video series which uses noisycharts to highlight the best of our data journalism across multiple platforms.
Explore adding noisycharts into Guardian’s audio reporting where appropriate (eg turn a chart of inflation into audio and play it in the news podcast for an episode about inflation).
