I’m planning on using random numbers to generate data that will allow me to find a single spoken word in a film as the basis for my experiments.

I selected film as a medium because, although I enjoy film, my enjoyment of it is definitely more recreational in nature. I wanted to pick a medium, starting from my generator, that was already on the peripheries of my comfort zone, so I decided to draw on my partner’s library as it has a few hundred films that vary in style and content.

Process

To find the single word, I will make 10 small polymer clay beads labelled 0 to 9. Each ball was weighed to the nearest gram to try to keep the beads as uniform as possible, to eliminate data skewing. I used a silver paint pen to mark the surface with a number, as this also ensured that the number could not be identified by any raised ridges or other defects in the polymer clay surface.

I will draw three numbers to select which film from the list (533, sorted in order of most recent download, to further randomise the list). Once the film is selected, numbers will be drawn to correspond with a timestamp of the film (in 1 second increments). One number will be drawn to select the hour of the film (beads will be removed from the pool as required), then two numbers for minutes, and then a following two for seconds. This second batch of numbers will correspond with a timestamp in the film.

Should the selected timestamp not include dialogue, the full batch (film number included) will be redrawn. I feel like authenticity is integral to this process.

Data

The resulting randomised numbers will follow the below format:
Film number: ###
Timestamp: #:##:##

I will also take a screenshot of the frame where the word is spoken, and will make a short video sample to demonstrate the spoken word in context (tone, volume, etc). I think that this will provide me with a rich selection of data to work from.

I did consider generating a number to find a specific frame, however, I felt that it would be overcomplicating the process. By finding a timestamp of one second, I am going to generate more data options (sound, language, tone, etc) than from a single frame.

To see the results, check out the Results page!