DATA MODEL

BACKGROUND

 Participants were asked to collect and describe three sounds of their everyday lives: one familiar, one unusual, and one meaningful. A short description of the sound and its location accompanied each submission. Lastly, participants were asked to use location categories and affective quality tags to label the sound and its place of origin.

THE PLAN

Here was my original data model sketch:

THE REALITY

Twenty people gathered and shared sounds of their everyday lives during the spring of 2020, amidst the coronavirus pandemic, telecommunication, social distancing, and a cumulative, noticeable change in our daily sonic environments.

This is a pilot project for working with data in crowd-sourced soundscaping. For a cursory analysis, here is what I have done so far:

1. quantitatively analyzed the qualities of sound by their three types (familiar, unusual, meaningful) using Tableau Public

2. qualitatively coded the written descriptions accompanying each submission for themes using NVivo

3. mapped the sounds using Echoes, an interactive geolocation app that allows users to hear sounds in a particular area

4. compiled sounds in Ableton Live to visualize waveforms

5. stitched together audio collages by the three categories in Ableton Live

As is clear from the contrast between plan and reality, I ran into trouble with ArcGIS and StoryMaps because of their limitations on uploading sound media. The rest, surprisingly, went pretty much according to plan!

 DATA COLLECTION

Here’s a glimpse from behind-the-scenes in TypeForm, the website I used to collect my crowd-sourced data.

 DATASET REFINING AND TIDYING

I connected TypeForm to my Google Drive to collect responses automatically in Google Sheets. As it was my first TypeForm rodeo, I’m not entirely sure what I did wrong, but the responses did not come out in question order, which was a problem for some of the metadata that went along with the sound clips.

I ended up going back through the responses and confirming the connections between submitted sound clips and their metadata (location, tags, time, date, and so on). My goal was to arrange the data in an order that made sense in my brain and allowed me to triple check the completeness of the submissions.

Lastly, I employed some data tidying strategies, namely organizing my data rows by single observation rather than participant. So, now each participant has three rows, corresponding to each of their submissions. I did run into some problems with the sound tag categories for location and qualities. I ended up making a column for each of the potential checkbox qualities and using 1s and 0s to numerically account for the attributes. The result is a somewhat unwieldy 57 sound segments with a range of qualities for each. (I’ll be thinking long and hard about I can more strategically structure the questions and logic map in TypeForm to gather responses that will be more tidy from the start).