Course materials for UCL module COMP0161
Your task is to sonify some data – that is, produce an audio representation of the data that captures (some of) its salient features in an intuitive or revealing way.
Some candidate data sources are given below, but you are welcome to choose something else – indeed, we very much encourage you to select a dataset that aligns with your own interests.
Ifat covered different approaches to sonification in her lecture of week 3, and we looked at a very simple example in that week’s tutorial. You may draw on that example, but you must do your own analysis and design – do not simply reproduce the (often arbitrary) choices from the tutorial.
You are required to submit a single fixed sound file, so you will probably find parameter-based sonification approaches more suited to the task. If you prefer to produce a model-based interactive sonification you may do so, but you will need to record a “walkthrough” that you can submit.
You may reuse audio code from the tutorials, or write your own, or use other libraries. You may make use of field recordings or found sounds or existing samples. If you want to build your whole sonification out of the Wilhelm Scream or Amen Break, you can. You may use most other tools for audio or music creation, provided the work you do with them is your own. (See the section on GenAI for more on this.) Some potential options are mentioned in the Resources section below, but you are not required to use any of them. If you are already familiar with a particular audio application or environment, feel free to use that.
In all cases, you must cite your sources in your report and make clear how you have used them.
You may use any data source that appeals to you, but you should look for something with sufficient variety to produce interesting results. Some possible candidate datasets from Kaggle are noted below, but you are welcome to ignore these and use something completely different. Cite and, where possible, include a link to your source(s) in your report.
A good starting point for figuring out your design is the Data Sonification Canvas by Sara Lenzi & Paolo Ciuccarelli (see also the paper about it). Their Data Sonification Archive includes many interesting sonification examples.
The Sonification Handbook is the “bible” of sonification, covering many aspects of the field. It is freely available in PDF form from that site.
There are many computational tools available for creating and manipulating sounds. A few are even dedicated to sonification:
(It is okay to use these, but you may find them limiting – they impose their own assumptions about what’s being done, which may not accord with yours.)
You will probably need an audio editor. If you don’t have anything that you’re already familiar with, Audacity is free, capable and very widely used. You probably won’t need a “Desktop Audio Workstation” or DAW, and if you do you probably know it already. But GarageBand is free on Mac & iOS, Cakewalk is free (up to a point) on Windows & Mac, and BandLab is free online.
Most coding environments will enable you to produce audio, perhaps with the installation of additional libraries. As we’ve seen in the tutorials, low level generation is reasonably straightforward in Python, and libraries such as Pedalboard implement useful signal processing functionality. Web Audio is built into modern browsers and accessible via JavaScript. Tone.js uses it to provide a handy framework for audio synthesis and control. Alternatively, p5.sound exposes Web Audio functionality for p5.js. If you’re happier with C++ then JUCE is a very popular cross-platform library (commercial, but free for non-commercial and educational uses). That may be a bit heavyweight for this coursework, though.
There are also plenty of dedicated audio & music coding environments, including:
TidalCycles is a nice livecoding environment that uses SuperCollider as its synth engine. But if you just want to give it a try you’re probably better off with Strudel, a JavaScript version that runs in your browser.
Freesound is a good source of audio samples. As always, remember to credit your sources.
This coursework falls broadly into UCL Category 2 with respect to use of GenAI: GenAI tools can be used in an assistive role. However, due to the nature of the coursework, we take an extremely narrow view of what constitutes legitimate assistance. Please carefully read the notes below, and if you are even slightly uncertain about the acceptability of any intended use of GenAI in your work, ask for clarification.
Guiding principle: both the sonification and your report explaining it must be strictly your own work.
As we will discuss in the final lectures of the module, questions of authorship in the context of algorithmic composition and generative media can be quite fraught. The author’s influence over the end product might be deliberately indirect. It is legitimate for that to be the case for your sonification – the content will necessarily be determined by the data to a significant extent.
Nevertheless, you must be able to demonstrate unambiguously that you are in fact the author. You must justify your choices and actively implement them. You must not delegate any of those choices to an “AI” agent.
Specifically:
IMPORTANT: You must provide a statement in your report explicitly documenting all uses of GenAI, specifying the systems and/or models used and what they were used for. This statement counts towards your page and word limits. It is not worth any marks in itself.
Your submission must consist of exactly 2 files, submitted together a single zip archive:
Length limits will be strictly enforced. Audio files and/or reports that exceed the stipulated length will be cut off at the limit and marked as if the remaining material had not been submitted. There is no penalty attached to the overrun itself, but your truncated coursework will likely be incomplete and lose marks as a result.
The report is tightly constrained and has a lot of ground to cover, so be succinct and to the point. Do not waste words on niceties.
A suggested report structure is given below. This is not a strict prescription – you may organise the material differently if you think doing so presents your work more clearly – but you do need to address all these points.
You may include figures and tables, but they must fit within the page limit. Ensure that all labels and captions are at a legible size – anything we can’t read will be treated as not having been included at all.
100 marks are available for this coursework, which will contribute 30% of your overall grade for the module. The marks are divided equally between the audio and the report, apportioned as follows:
It should be clear that there is no single correct answer to this task. We are not expecting you to get it “right”, so please don’t try. The point of the exercise is to see what choices you make, how you follow through, and what you discover along the way.