What is a ‘machine’ anyway? Help us describe them

Written by Mia RidgeOctober 27, 2020Comments: 0

Help us describe machines: contribute to our lists of ‘machine’ words on Zooniverse.

Our first crowdsourcing tasks, which asked people to look for industrial accidents described in newspaper articles, raised some really important questions. Why is the language used to describe some accidents so grisly? Are accidents close to the place of publication written about differently than accidents far away? 

They also raised a more fundamental question: what do we mean by ‘machine’? Our working definition for the first task was that machines are ‘devices or equipment not powered by people or animals’. That excluded bicycles, horse-and-cart vehicles and manual equipment, and allowed us to focus on the impact of mechanisation and industrialisation, especially in workplaces outside the home. However, those first tasks reinforced the extent to which our current understanding of what counts as a ‘machine’ reflects our times, not our research period. 

So what did writers (and presumably readers) of 19th British newspapers understand ‘machines’ to be? We’re launching a new crowdsourcing task called ‘What’s that machine? Describe it!‘ to find out. Head over to our Zooniverse site to take part, or read on to find out more.

What's that machine? Describe it!

Help us create a lexicon for machines

Our new task goes back to basics and asks: what did people in the 19th century mean when they wrote ‘machine’? We’re asking the public to help us create a lexicon (a ‘dictionary’ of words) of machines by reading articles and recording the machines mentioned. One result will be a list of ‘hyponyms’ – words for more specific types of machines – that can be used to scale up queries across our corpora. This builds on work elsewhere in the project that looks at different meanings given for ‘machine’ in the Oxford English Dictionary (OED).

Our linguists call it the ‘anaphora task’. In everyday language, anaphora is when you talk about something and then you refer back to it. An example from an article might be ‘try our silent sewing machine. This machine will impress you’, where ‘this’ refers to the silent sewing machine. ‘Anaphora resolution’ is the process of linking ‘this machine’ to ‘silent sewing machine’ automatically; the results of this task might help computers to learn how to find anaphoras in text automatically.

How does the task work?

We built scripts to search our corpus of newspapers published in English in the UK during our period of interest (broadly, the 19th century plus a bit either side) for articles that use the phrase ‘this machine’. The scripts highlight the ‘machine’ phrase, and crop the image to a few hundred words either side of it to help people focus on a specific reference. We then built a public crowdsourcing task on the Zooniverse platform based on the research question we’d agreed over many discussions. The task asks participants to look at the highlighted phrase in the article and type in words that name or define the ‘machine’ in question.

We ask people to use words as close as possible to those written in the text. The workflow is designed to weed out unsuitable images so that only images with readable text about a physical machine go into the next stage. We’ll check for interesting ‘disagreements’ between classifications, and explore the differences in definitions between participants as the results come in.

Join in!

You can try the task for yourself on Zooniverse. We’d love your feedback and ideas in the forum, on twitter or email.

The work discussed in this post includes contributions by Kaspar Beelen, Sarah Gibson, Barbara McGillivray and Mia Ridge, with additional support from Kasra Hosseini, Federico Nanni and Daniel Wilson. Thanks to Ruth Ahnert, Graham Jevon, Giorgia Tolfo, David Hughes and Bruce Wyman for usability reviews.

Our Partners

Our Funders