65. Digital Death Trip: Finding random olden days town tragedies with Trove, code, and robot_TMBTP

In this experimental episode of This Must Be The Place Elizabeth pilots a computer program named Digital Death Trip, or alternatively Trove Town Tragedy, coded by Elizabeth’s sister Sarah. The program selects a random location in Victoria, then a random ‘tragedy’ from that location using the National Library of Australia’s Trove archive of digitized newspapers, and the Trove API. The olden days articles retrieved are mostly from between the 1860s and 1950s. The program uses ‘Speak’ to read out articles in a robot voice. It’s art meets historical geography meets Perfect Match robot. Death Trip is a reference to 1973 book Wisconsin Death Trip, a collection by Michael Lesy of photographs and news articles from two small superficially comfortable Wisconsin towns in the late 19th century. By picking darker articles – suicides, epidemics, bankruptcies, murders, tramp armies – the effect was “constantly repeated themes”, with fragments illustrating each another and the disruptions of their times. Elizabeth wondered if something broadly similar could be done with Australia’s Trove. Trove’s API allows you to use a key to search and use results automatically, to create your own application (one example is the Trove Penguin Bot). Much of Trove is open and crowd-sourced - anyone can search, or correct machine-read text, or create lists. While it doesn’t yet have specific location data, experiments like this bring more of a spatial dimension. What the Death Trip code does is randomly select a place name in Victoria, combine this with the key word “tragedy”, search Trove, shortlist and read out headlines, and compile a case file on a random article. Depending what comes back, the idea is to investigate the selected town tragedy. This could be an excuse to go on trips to visit Victorian towns, perhaps some with swimming pools. While the code could search for any kind of event or word, this project is specifically interested in the word ‘tragedy’. Tragedy was a common news headline partly because covering Coroner’s courts and police beats was a convention of the emerging popular newspaper industry of the 19th century. The articles the bot retrieves are not just any tragedies, but those tied to a place, usually with ‘the’ definite article. Nearly every Australian town seemed to have had incidents headlined The [insert Town Name] Tragedy. These included forces of nature – drowning and fires, and what we might now call negligence involving workplace machinery. Car accidents, as we now know them, were at first ‘tragedies’. And violence: it covered suicides, murders, and murder-suicides particularly in the context of the home. Naming things ‘The’ town tragedy seemed to suggest such things couldn’t and haven’t happened in that place, and that if they did they would leave a lasting impact. One thing the experiment shows is these things have happened, usually more than once, and they are forgotten, and happen again. Tragedy seemed then, as now, to mean the line between things we want to try to understand or control, and those we do not. Some of these change, and some don’t (notably, some of the language of domestic murder suicides). As part of this project, Elizabeth has so far corrected and tagged about 1,100 tragedies. Words like farm, ‘pea-rifle’, and ‘quarrel’ have new connotations. In this test, Elizabeth and Sarah run through the concept, then the mechanics and mishaps of running the code, and see which towns and tragedies the bot picks. As it turns out, those selected are from 2 irrigation towns in Victoria: The Tatura Tragedy 1905; and The Quantong Tragedy 1894. As a follow up, Elizabeth will use the results to find out more about the incidents and their contexts in time and place. Trove Town tragedy list: trove.nla.gov.au/newspaper/result…trovetowntragedy Github project: github.com/SarahTaylorProject/trove_experiment